HK1193471B - Identifying and measuring reticulocytes - Google Patents

Description

Identifying and measuring reticulocytes

Cross reference to related applications

The present application claims priority based on 35u.s.c. § 119(e) to the following U.S. provisional patent applications: 61/510,614, filed on 7/22/2011; and 61/510,710, filed on 7/22/2011. The entire contents of each of the aforementioned applications are hereby incorporated by reference.

Technical Field

The present disclosure relates to identifying, enumerating, and measuring properties of reticulocytes in a biological sample comprising blood and a quality control complex designated to mimic blood.

Background

Reticulocytes are immature red blood cells that are generally characterized by elevated levels of RNA (ribonucleic acid) compared to mature red blood cells. Reticulocyte counts performed on blood samples can provide important information about how quickly the patient's bone marrow is producing red blood cells. Thus, the number of reticulocytes in the patient's blood is used as an important diagnostic indicator, usually a parameter of the complete blood count. For example, a high reticulocyte count may indicate possible internal trauma, blood loss, or some type of anemia. A low reticulocyte count may indicate a certain nutrient deficiency (e.g., iron deficiency) or disease affecting the bone marrow like cancer. Disclosed herein are methods and systems for automated measurement of cellular samples, including identification, quantification, and constituent measurement of biological samples including blood, as well as reticulocytes in quality control complexes designated to mimic blood.

Disclosure of Invention

The methods and systems disclosed herein allow for the automatic identification and quantification of reticulocytes in a blood sample using spectral imaging. This identification and quantification is based on images of one or more cells, which images are obtained at different light wavelengths. In particular, the disclosed methods and systems may be used to identify and measure properties of reticulocytes in a blood sample drawn from a patient. The properties may include a variety of parameters related to the shape and appearance of the reticulocytes, including area, perimeter, circularity, optical density, and spectral properties at different wavelengths. The reticulocytes can be selectively displayed to the technician (e.g., highlighted to distinguish them from mature red blood cells). The diagnostic information obtained from the reticulocytes can also be displayed to a technician for further consideration (e.g., to determine the condition that the patient is likely to be experiencing and to select an appropriate course of treatment) and/or can be communicated to other automated devices, stored in electronic medical records, and sent to other health care professionals.

In general, in a first aspect, the disclosure features a method of identifying reticulocytes in a blood sample deposited on a substrate, the method comprising: illuminating the sample with incident light at two different wavelengths, acquiring a two-dimensional image of the sample corresponding to a first wavelength, and acquiring a two-dimensional image of the sample corresponding to a second wavelength; analyzing the image to identify a set of representative red blood cells; determining the area of each red blood cell in the set; determining a color value of each red blood cell in the set; and for each red blood cell in the set, identifying the red blood cell as a reticulocyte if the area of the red blood cell exceeds an area cutoff value and the color value of the red blood cell is less than a color cutoff value, wherein the color value of each red blood cell comprises a difference between optical density values of the red blood cell at two illumination wavelengths.

Embodiments of the method may include any one or more of the features disclosed herein.

One of the wavelengths may be between 400nm (nanometers) and 457nm, and the other wavelength may be between 575nm and 600 nm.

Determining the color value of each of the identified red blood cells may include, for each red blood cell: determining a set of pixels associated with the cell; determining, for each of the set of pixels, an optical density corresponding to the first wavelength; determining, for each of the set of pixels, an optical density corresponding to the second wavelength; determining, for the set of pixels, an average optical density corresponding to the first wavelength; determining, for the set of pixels, an average optical density corresponding to the second wavelength; and calculating the difference between the average optical densities to determine a color value for the cell.

Determining the area of each red blood cell in the set may include, for each red blood cell, determining a set of pixels associated with the cell, and determining the area of the cell by counting the number of pixels in the set. Determining the area of each red blood cell in the set may include, for each red blood cell, determining a set of pixels associated with the cell, determining a polygon surrounding the set of pixels, and determining the area of the cell by calculating the area of the polygon.

The color cutoff value may be based on a percentile of the distribution of color values of the red blood cells. The percentile may correspond to 70% within the distribution of color values of the red blood cells. The color cutoff value may correspond to a sum of the percentile and the color offset value. The method may include determining percentile and color offset values based on a set of training data for which the number of reticulocytes is known. The area cutoff value may be based on a percentile of the distribution of the areas of the red blood cells. The percentile may correspond to 20% of the area distribution of the red blood cells. The area cutoff value may correspond to a sum of the percentile and the area offset value. The method may include determining percentile and area offset values based on a set of training data for which the number of reticulocytes is known.

The method may include excluding the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at one of the two wavelengths is greater than a cutoff value. The method may include excluding the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at wavelengths other than the two wavelengths is greater than a cutoff value.

The method may comprise, for each red blood cell identified as a reticulocyte, determining a volume of the reticulocyte. The method may include determining a volume of the reticulocytes from the integrated optical density of the reticulocytes corresponding to the plurality of illumination wavelengths. The method may include determining a mean reticulocyte volume parameter for the sample.

The method may include, for each red blood cell identified as a reticulocyte, determining a hemoglobin content of the reticulocyte. The method can include determining a hemoglobin content of the reticulocytes from a weighted combination of an area of the reticulocytes, a volume of the reticulocytes corresponding to the plurality of illumination wavelengths, and a combined optical density of the reticulocytes corresponding to the plurality of illumination wavelengths. The method may include determining a mean reticulocyte hemoglobin value for the sample.

Embodiments of the method may also include any of the other features or steps disclosed herein, in any combination, as appropriate.

In another aspect, the disclosure features a system for identifying reticulocytes in a blood sample deposited on a substrate, the system including a light source configured to illuminate the sample with incident light at two different wavelengths; a detector configured to acquire a two-dimensional image of the sample corresponding to a first wavelength and to acquire a two-dimensional image of the sample corresponding to a second wavelength; and an electronic processor configured to analyze the image to identify a set of representative red blood cells; determining the area of each red blood cell in the set; determining a color value of each red blood cell in the set; and for each red blood cell in the set, identifying the red blood cell as a reticulocyte if the area of the red blood cell exceeds an area cutoff value and the color value of the red blood cell is less than a color cutoff value, wherein the electronic processor is configured to determine the color value of each red blood cell based on a difference between optical density values of the red blood cell at two illumination wavelengths.

Embodiments of the system may include any one or more of the following features.

The first wavelength may be between 400nm and 457nm and the second wavelength may be between 575nm and 600 nm.

For each red blood cell, the electronic processor can be configured to determine a color value of the cell by: determining a set of pixels associated with the cell; determining, for each of the set of pixels, an optical density corresponding to the first wavelength; determining, for each of the set of pixels, an optical density corresponding to the second wavelength; determining, for the set of pixels, an average optical density corresponding to the first wavelength; determining, for the set of pixels, an average optical density corresponding to the second wavelength; and calculating the difference between the average optical densities to determine a color value for the cell.

For each red blood cell in the set, the electronic processor can be configured to determine the area of the cell by: a set of pixels associated with the cell is determined, and the area of the cell is determined by counting the number of pixels in the set. For each red blood cell, the electronic processor can be configured to determine the area of the cell by: the method includes determining a set of pixels associated with the cell, determining a polygon surrounding the set of pixels, and determining an area of the cell by calculating an area of the polygon.

The electronic processor can be configured to determine a color cutoff value based on a percentile of a distribution of color values of the red blood cells. The percentile may correspond to 70% within the distribution of color values of the red blood cells. The electronic processor can be configured to determine the color cutoff value as a sum of the percentile and a color offset value. The electronic processor can be configured to determine the percentile and the color offset value based on a set of training data in which the number of reticulocytes is known. The electronic processor can be configured to determine an area cutoff value based on a percentile of the distribution of areas of the red blood cells. The percentile may correspond to 20% of the area distribution of the red blood cells. The electronic processor can be configured to determine the area cutoff value as a sum of the percentile and an area offset value. The electronic processor can be configured to determine the percentile and the area offset value based on a set of training data in which the number of reticulocytes is known.

The electronic processor can be configured to exclude the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at one of the two wavelengths is greater than a cutoff value. The electronic processor can be configured to exclude the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at wavelengths other than the two wavelengths is greater than a cutoff value.

For each red blood cell identified as a reticulocyte, the electronic processor can be configured to determine a volume of the reticulocyte. The electronic processor can be configured to determine a volume of the reticulocyte based on the integrated optical densities of the reticulocyte corresponding to the plurality of illumination wavelengths. The electronic processor can be configured to determine an average reticulocyte volume parameter for the sample.

For each red blood cell identified as a reticulocyte, the electronic processor can be configured to determine a hemoglobin content of the reticulocyte. The electronic processor can be configured to determine a hemoglobin content of the reticulocyte based on a weighted combination of an area of the reticulocyte, a volume of the reticulocyte corresponding to the plurality of illumination wavelengths, and a combined optical density of the reticulocyte corresponding to the plurality of illumination wavelengths. The electronic processor can be configured to determine a mean reticulocyte hemoglobin value for the sample.

Embodiments of the system may also include any of the other features or steps disclosed herein, in any combination, as appropriate.

Unless defined otherwise, all scientific and technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of the present invention, suitable methods and materials are described below. All publications, patent applications, patents, and other references mentioned herein are incorporated by reference in their entirety. In case of conflict, the present specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and not intended to be limiting.

The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features and advantages will be apparent from the description and drawings, and from the claims.

Drawings

FIG. 1 is a schematic diagram of a system for identifying, quantifying, and measuring characteristics of reticulocytes in a biological sample;

FIG. 2 is a flow chart showing a series of steps for identifying reticulocytes in a sample;

FIG. 3A is a graph illustrating a color cutoff for reticulocyte recognition among red blood cells in a sample;

FIG. 3B is a graph illustrating an area cutoff value for reticulocyte identification among red blood cells in a sample;

FIG. 4 is a schematic diagram showing a series of steps for selecting a representative set of red blood cells from one or more images of a sample;

FIG. 5 is a schematic diagram of a cell showing the cell boundaries;

FIG. 6 is a schematic diagram showing two cells and a convex hull determined for each cell;

FIG. 7 is a schematic diagram of a cell showing the variation of optical density among cell pixels;

FIG. 8 is a schematic diagram of an automated sample processing system; and

FIG. 9 is a schematic diagram of a computing system that measures the volume, composition, and other characteristics of reticulocytes.

Like reference symbols in the various drawings indicate like elements.

Detailed Description

Reticulocytes are immature red blood cells that are generally characterized by elevated levels of RNA compared to mature red blood cells. Reticulocyte counts performed on blood samples can provide important information about how quickly the patient's bone marrow is producing red blood cells. Thus, the number of reticulocytes in the patient's blood is used as an important diagnostic indicator, usually a parameter of the complete blood count. For example, a high reticulocyte count may indicate possible internal trauma, blood loss, or some type of anemia. A low reticulocyte count may indicate a certain nutrient deficiency (e.g., iron deficiency) or disease affecting the bone marrow like cancer. Disclosed herein are methods and systems for automated measurement of cell samples, including identification, quantification, and compositional measurement of reticulocytes in the sample. In particular, the methods and systems disclosed herein can be used to measure samples and identify reticulocytes in an automated manner in a quality control composition intended for use on an automated laboratory analyzer or in a variety of different other samples including samples taken from animals. The methods and systems disclosed herein allow for high throughput, fully automated analysis of a variety of biological samples drawn from a patient without the need for special stains or other contrast agents specifically designed for reticulocyte identification.

An automated system for preparing samples for analysis is described in more detail below. Once the sample is prepared, it is transported to an automated measurement system. The measurement system acquires one or more two-dimensional images of cells in the sample and uses the images to identify reticulocytes among the red blood cells and other forming elements of the blood in the images, and in some embodiments, measures a quantity related to the identified reticulocytes such as reticulocyte volume, area, perimeter, optical density, and/or a quantity of one or more components (e.g., hemoglobin) in the identified reticulocytes. These quantities are determined from information derived from an image of the reticulocyte obtained by directing incident light onto the reticulocyte and then detecting the portion of the incident light that is transmitted through or reflected from the reticulocyte. Each image is a two-dimensional image in which the individual pixel intensity values within the image correspond to the amount of transmitted or reflected light that emerges from the spatial location on the reticulocyte corresponding to the pixel.

General considerations

Reticulocytes typically include a network of ribosomal RNAs that are present in the reticulocyte at a greater concentration than other types of cells, including mature red blood cells. When certain stains, such as sky blue B and methylene blue, are applied to the reticulocytes, the stains bind to the RNA. Upon examination under the microscope, the stained RNA was readily visible, while the reticulocytes appeared slightly bluish and larger than the mature red blood cells. The amount of stained RNA is one criterion that can be used to distinguish reticulocytes from other cell types.

Automated sample preparation systems may apply a variety of different stains to a blood sample. While certain conventional techniques for identifying reticulocytes rely on stains that are specific for binding cellular RNA or other mechanisms specifically intended for reticulocytes, the methods and systems disclosed herein advantageously rely on stains that are capable of automatically detecting and measuring all other parameters of a complete blood count, including the romanofsky (Romanowsky) stain. Thus, a stain targeting both cellular RNA and DNA can be applied to the reticulocyte, which can be distinguished from other cell types due to the staining network of ribosomal RNA described above. More generally, cells comprising red blood cells (e.g., mature red blood cells, reticulocytes) are typically prepared (as part of a blood sample) by applying one or more staining agents to the blood sample. The stain binds to the cytoplasm and serves as a marker for the cytoplasm in the cell image. When incident light is irradiated on a cell, the staining agent absorbs a part of the incident light; the amount of absorption at a particular location in the sample depends on the amount of stain present in that location.

In addition to identifying reticulocytes, the methods and systems disclosed herein can be used to determine various properties of the identified reticulocytes. For example, by decoupling the estimate of cell thickness from the absorption effects of the cell components (e.g., hemoglobin) and the locally varying concentrations of various applied stains, the reticulocyte volume can be determined from information derived from a two-dimensional reticulocyte image. To achieve such decoupling, the pixel intensities are scaled by the maximum pixel intensity of each cell. As described further below, the reticulocyte volume calculation may be based on a weighted combination of the optical density value and the reticulocyte area for each color of illumination light used to acquire the reticulocyte image. The process described herein can be repeated for each member of a set of reticulocytes selected for the volume measurement, and the results used to calculate the average reticulocyte volume of the sample. As described further below, the number of reticulocyte components, such as the hemoglobin content of the reticulocytes, can also be determined from calculations based on the optical density values for each color of illumination light used to obtain the reticulocyte image.

For purposes of the following discussion, it is assumed that at least two stains are applied to the sample: eosin and sky blue. However, the methods and systems disclosed herein are not limited to applying only two stains or only eosin and sky blue. Rather, the method and system are capable of measuring samples to which fewer stains (e.g., one stain) or more stains (e.g., red stains including eosin and stains including sky blue and methylene blue, three or more stains, four or more stains, five or more stains) have been applied.

Fig. 1 shows a schematic diagram of a system 100 (which may be part of a larger sample processing and analysis system) for automatically identifying and measuring reticulocytes in a blood sample. The subsystem 100 includes an illumination source 102, a detector 106, and an electronic control system 108. The electronic control system 108 may include a display 110, a human interface unit 112, and an electronic processor 114. The electronic control system 108 is connected to the illumination source 100 and the detector 106 via control lines 120 and 122, respectively.

Assuming that a sample has been prepared for analysis (as discussed further below), prepared sample 104 (e.g., a blood sample that is deposited on a microscope slide, then fixed, stained, washed, and dried) is automatically placed in proximity to source 102. Source 102 directs incident light 116 toward sample 104. A portion of the incident light passes through the sample 104 as transmitted light 118, which is detected by the detector 106. The transmitted light 118 forms an image of the sample 104 on the active surface of the detector 106; the detector 106 captures the image and then sends the image information to the electronic control system 108. Generally, the electronic control system 108 directs the source 102 to produce incident light 116 and directs the detector 106 to detect an image of the sample 104. When the detector 106 acquires an image of the sample 104 from the transmitted light 118, the control system 108 may instruct the source 102 to use a different illumination wavelength.

The process discussed above may be repeated for multiple images of sample 104, if desired. Prior to acquiring a new image, electronic control system 108 may adjust the wavelength of incident light 116 generated by source 102. In this way, each image of the sample 104 may correspond to a different wavelength of incident light 116, and thus a different wavelength of transmitted light 118. This process is repeated until at least sufficient information is obtained to perform an accurate identification of red blood cells including reticulocyte candidates and/or to determine properties associated with the identified red blood cells and/or reticulocytes.

Typically, the amount of information that yields accurate identification and/or quantification of reticulocytes is determined during the calibration process. For example, a calibration process may be used to determine that accurate reticulocyte identification and quantification may be achieved when the number of sample images obtained is greater than or equal to the number of spectral contributors (e.g., absorbers) that are a factor in analyzing the sample. As an example, for a prepared sample containing reticulocytes as a naturally occurring absorber, and red blood cells as eosin and sky blue applied stains (for a total of three spectral contributors), the system 100 may continue to acquire sample images until images are obtained at a minimum of three different wavelengths. Additional images corresponding to further different wavelengths may also be obtained and used in reticulocyte identification and quantification.

The illumination source 102 may include one source or a plurality of the same or different sources to direct incident light onto the sample. In some embodiments, source 102 may include a plurality of light-emitting elements such as: diodes (LEDs), laser diodes, fluorescent lamps, incandescent lamps, and/or flash lamps. For example, source 102 may include four LEDs having output wavelengths in the red, yellow, green, and blue regions of the electromagnetic spectrum (e.g., 635, 598, 525, and 415 nm), respectively, or more generally, about 620 to 750nm (red), about 575 to 600nm (yellow), about 525 to 570nm (green), and about 400 to 475nm (blue). In certain embodiments, the source 102 may include one or more laser sources. Instead of having multiple light emitting sources, in other embodiments, the source 102 may comprise a single broadband light emitting source that may be configured to alter its output wavelength (e.g., under the control of the electronic control system 108). For example, source 102 may include a broadband source (e.g., an incandescent lamp) coupled with a configurable filtering system (e.g., a plurality of mechanically adjustable filters, and/or a liquid crystal-based electronically adjustable filter) that produces a variable output spectrum under control of system 108. In general, the source 102 does not output the illumination light 116 at a single wavelength, but rather at a band of wavelengths around a center wavelength (e.g., the wavelength of maximum intensity in the band of wavelengths). When the discussion herein refers to the wavelength of the illumination light 116, such reference refers to the center wavelength of the illumination band.

The detector 106 may include a variety of different types of detectors. In some embodiments, detector 106 comprises a Charge Coupled Device (CCD). In some embodiments, the detector 106 may include a photodiode (e.g., a two-dimensional photodiode array). In some embodiments, the detector 106 may include other photosensitive elements such as CMOS based sensors and/or photomultipliers. As described above in connection with the source 102, the detector 106 may also include one or more filtering elements. In some embodiments, sample images corresponding to different waveforms are obtained by: illuminating light 116 having a relatively broad distribution of wavelengths is directed onto the sample 104, and the transmitted light 118 is then filtered to select only the portion of the transmitted light corresponding to the small band of wavelengths. Filtering may be performed on one or both of the excitation side (e.g., in source 102) and the detection side (e.g., in detector 106) to ensure that each of the images obtained using detector 106 corresponds to a particular distribution of light wavelengths having a particular center wavelength.

General reticulocyte recognition method

The systems and methods disclosed herein acquire images of cells in a sample (e.g., a blood sample deposited on a microscope slide and subsequently prepared for automated microscope imaging using fixatives and stains), and identify and quantify reticulocytes from among the cell images. Fig. 2 shows a flow chart 200 of various steps included in such an image to identify reticulocytes. In a first step 202, a representative set of red blood cells is identified from among all white blood cells, red blood cells, platelets, and cellular and non-cellular artifacts in one or more acquired sample images. In a next step 204, reticulocytes are identified among the set of representative red blood cells. Then, in optional step 206, reticulocyte characteristics and metrics, such as reticulocyte volume, the number of reticulocyte components, such as hemoglobin, and reticulocyte geometry parameters, may be determined from the reticulocytes located in step 204. The process terminates at step 208. Each step in flowchart 200 is described in more detail below.

(i) Reticular cell recognition

As discussed above, the first step in identifying reticulocytes in a blood sample is to identify a representative set of red blood cells (of which the reticulocytes form a subset) among all white blood cells, red blood cells, platelets, and cellular and non-cellular artifacts in one or more of the acquired sample images. This process will be described in subheading "(ii) Identifying a set of representative red blood cells"discussed in more detail below, and may include one or more feature measurement and artifact exclusion techniques that identify red blood cells from an image of a sample containing red blood cells, white blood cells, and platelets. After identifying the representative set of red blood cells, reticulocytes are identified among the representative red blood cells by further analyzing each member of the set of red blood cells. Because reticulocytes are non-mature red blood cells, a method of distinguishing red blood cells in a sample from other types of cells (e.g., white blood cells, platelets) can identify both reticulocytes and mature red blood cells together as the set of representative red blood cells. Localizing reticulocytes involves distinguishing reticulocytes from other more mature red blood cells in the representative group.

In some embodiments, reticulocytes may be localized in this representative group, sometimes by means of a colorant or stain (e.g., using a combination of fixative, stain, and wash agents disclosed in U.S. patent application No. 13/526,164 filed 6/18/2012, the entire contents of which are incorporated herein by reference) according to their absorption spectra. For example, because reticulocytes contain a network of ribosomal RNAs, a stain that binds to nucleic acids can be used to identify reticulocytes, as RNA is typically present in elevated concentrations within the reticulocyte.

Identifying reticulocytes among the set of representative red blood cells can be determined from images obtained in one or more wavelength regions of the electromagnetic spectrum based on the optical densities of the individual cells in the set. For example, in some embodiments, reticulocytes can be localized by examining the optical density of each representative red blood cell in a sample image obtained in the yellow region of the spectrum. Reticulocytes typically have a stronger absorption than mature red blood cells in the yellow region because the nucleic acid stain applied to red blood cells has an absorption peak in the yellow region of the spectrum. For example, Romanofsky stain is often used to label RNA of red blood cells (including reticulocytes) in a sample.

In certain embodiments, reticulocytes can be localized by examining the optical density of each representative red blood cell in a sample image obtained in the blue region of the spectrum. Typically, the blue region of the spectrum ranges from 400nm to 470nm for the purpose of identifying reticulocytes. Reticulocytes generally have a weaker uptake in the blue region than mature red blood cells because hemoglobin in the reticulocytes has not yet been completely formed. Each representative red blood cell can be classified as either a reticulocyte or a mature red blood cell by comparing the integrated optical density of each representative red blood cell in the blue region to a range of optical density values known to be associated with reticulocytes (e.g., based on flow cytometry analysis of a standardized sample).

In certain embodiments, the optical density of red blood cells in multiple spectral regions can be used to identify reticulocytes. For example, reticulocytes can be identified by comparing the optical density of each representative red blood cell in images obtained in the blue and yellow regions of the spectrum, respectively (e.g., images derived from cells obtained with blue and yellow illumination wavelengths). Reticulocyte identification may be aided by statistical analysis for the exclusion of imaging artifacts.

The first step in determining whether a particular representative red blood cell is a reticulocyte is to determine the color of the cell (C). For example, the color of a cell can be defined as:

C=OD_mean(b)–OD_mean(y)（1）

wherein OD_mean(b) Is the average optical density, and OD, of the entire pixel group corresponding to the red blood cells using the illumination wavelength in the blue region of the spectrum_mean(y) is the average optical density of the entire pixel set corresponding to red blood cells using the illumination wavelength in the yellow region of the spectrum. The determination of the pixel sets corresponding to the red blood cells is performed when the red blood cells are individually identified in the sample image, as will be discussed later.

The second step of determining whether a particular representative red blood cell is a reticulocyte is to determine the area (a) of the cell from the set of pixels corresponding to the red blood cell. As discussed further below, the cell area based on a set of pixels can be determined in a variety of ways. In some embodiments, the cell area is automatically determined upon identifying the set of pixels.

The color C and area a of the cell can then be used to determine whether the cell is a reticulocyte by comparing the values of these parameters to a cutoff value based on a statistical distribution of the values of these parameters throughout the representative population of red blood cells. A particular cell can be identified as a reticulocyte if the following equation holds true:

C<C_cutand A>A_cut（2）

Wherein C is_cutIs the color cutoff value, and A_cutIs the area cut-off value. Both the color and area cut-off values generally correspond to color and area among representative red blood cellsA particular percentile mark within the distribution of products. Since reticulocytes generally do not absorb as strongly as mature red blood cells in the blue region of the spectrum, the color C is less than the color cutoff value C_cutThe cells of (a) may be reticulocytes. Furthermore, since reticulocytes are larger than mature red blood cells, the area a is greater than the area cut-off a_cutThe cells of (a) may be reticulocytes. By identifying only cells satisfying two conditions as reticulocytes, reticulocytes can be localized within a larger representative group of red blood cells with high accuracy.

In some embodiments, only one of the conditions in equation (2) may be used to identify reticulocytes. For example, cells with a color value less than the color cutoff value can be identified as reticulocytes (i.e., C)<C_cut) Regardless of the area of the cell. As another example, cells having an area greater than the area cut-off may be identified as reticulocytes (i.e., A)>A_cut) Regardless of the color value of the cell.

As discussed above, the color and area cutoff values generally correspond to a particular percentile within the distribution of color and area among the representative red blood cells. By using percentiles (e.g., rather than absolute values), the methods disclosed herein may be used in samples where the staining protocol varies and/or in samples where the cell size naturally varies from the expected cell size (e.g., as seen in cell size typically varies from patient to patient). For example, the method can be used to locate reticulocytes among a representative set that is, on average, larger than red blood cells from other patients.

In some embodiments, the distribution percentile is adjusted using an offset value to arrive at a color and area cutoff value. For example, the color and area cutoff values may be expressed as:

C_cut=C_%+C_offset（3）

A_cut=A_%+A_offset（4）

wherein C is_%And A_%Color and area percentiles, respectively, and C_offsetAnd A_offsetColor and area offset values, respectively. FIG. 3A is a graph showing a color cutoff C for localizing reticulocytes in the set of representative red blood cells_cutA histogram of (a). In fig. 3A, a distribution 390 of the values of the color parameter C among all representative red blood cells is shown. Also shown within the distribution is a color percentile mark C_%. Since reticulocytes have a color value that is generally less than mature red blood cells, the color shift value C is_offsetIs negative, giving a color cutoff C less than the color percentile mark_cut. The shaded region 392 of the distribution corresponds to red blood cells identified as reticulocytes according to the color cutoff value.

FIG. 3B is a graph showing an area cut-off A for localizing reticulocytes in the representative set of red blood cells_cutA histogram of (a). In fig. 3B, the distribution 395 of the area among all representative red blood cells is shown. Also shown in this distribution is an area percentile mark A_%. Since reticulocytes have an area generally larger than mature red blood cells, the area offset value A_offsetIs a positive value, and an area cut-off value A greater than the area percentile mark is obtained_cut. The shaded region 397 of the distribution corresponds to red blood cells identified as reticulocytes by the area cutoff value.

Specific values for color and area percentiles and shifts can be determined by analyzing a training data set containing samples for which populations of mature red blood cells and reticulocytes are known. The training data set typically includes hundreds or thousands of samples; mature red blood cell and reticulocyte populations can be measured in these samples either manually by a laboratory technician or using a suitably calibrated flow cytometer system. The training set is then analyzed using the methods and systems disclosed herein, and the percentile and offset values are selected to best match the experimental analysis results to known measurements from the training data set.

As an example, in some embodiments, the color percentile mark is 70% and the color offset value is-0.09. More generally, the color percentile mark is 80% or less (e.g., 60% or less, 50% or less, 40% or less). Typically, but not always, if the color percentile mark is less than 70%, the color offset value is greater than-0.09.

As another example, in some embodiments, the color percentile mark is 20% and the color offset value is 12.4. More generally, the color percentile mark is 10% or greater (e.g., 30% or greater, 40% or greater, 50% or greater, 60% or greater). Typically, but not always, if the color percentile mark is greater than 20%, the color offset value is less than 12.4.

In some embodiments, one or more artifact-elimination steps are optionally performed to further refine the identification of reticulocytes from the set of representative red blood cells. These artifact removal steps may be performed before, after, or both before and after the comparisons described above for the color and area cutoffs. As an example, an artifact removal step based on statistical analysis of the optical density of pixels at different wavelengths corresponding to a particular representative cell may be used to remove cells as artifacts instead of cells. Optical density at multiple wavelengths (e.g., illumination wavelengths) may be used for artifact rejection. As an example, blue, yellow, and green wavelengths have been found to be useful for this purpose.

In general, cells having a group of related pixels with a large standard deviation of the optical density of the pixels may be classified as artifacts rather than cells, as a high standard deviation may indicate an unusual cell morphology, overlapping cells, and/or poor cell preservation. Such cells are generally unsuitable for further analysis. As an example, a cell of a group of related pixels can be excluded as an artifact if the following holds:

σ_OD(y)>σ_yor σ_OD(g)>σ_gOr σ_OD(b)>σ_b（5）

Wherein sigma_OD(y)、σ_OD(g) And σ_OD(b) Standard deviation of optical density of the set of pixels corresponding to images acquired with yellow, green and blue illumination wavelengths, respectively, and σ_y、σ_gAnd σ_bIs the standard deviation cutoff. In general, the appropriate standard deviation cutoff can be determined from analysis of training data as described above and/or from historical data (e.g., data from previously analyzed cell images). In some embodiments, for example, the appropriate standard deviation cutoff value is σ_y=0.1，σ_g=0.099, and σ_b= 0.058. These values can be varied according to the nature of the sample, the training protocol, and the imaging conditions. As an example, in certain embodiments, the appropriate standard deviation cutoff value is σ_y=0.079，σ_g=0.085, and σ_b=0.060。

The reticulocyte identification embodiments described above can also be used to identify reticulocytes in a quality control complex intended for use on an automated sample measurement system such as the system disclosed herein. Quality control compositions typically include various types of preserved mammalian blood cells designed to simulate a whole blood sample when processed on an automated system. The results of the analysis of these samples can, in turn, be used to assess the performance of automated systems, such as the accuracy and reproducibility of the system in identifying reticulocytes. In certain quality control complexes, the preservation process significantly reduces the size of reticulocytes in the control sample as compared to reticulocytes in the patient's blood sample; in some cases, the reticulocytes in such complexes are smaller than the mature red blood cell fraction. Thus, reticulocyte recognition embodiments that rely on the color analysis described above without using an area analysis can be used to accurately detect reticulocytes within certain quality control complexes.

(ii) Identifying a set of representative red blood cells

Returning to fig. 2, a first step 202 in flowchart 200 involves identifying a representative set of red blood cells from among the cells in an image of a sample obtained using the systems and methods described herein. The reticulocyte identification in step 204 occurs after the set of representative red blood cells has been identified from the image of the sample; thus, the set of representative red blood cells typically includes mature red blood cells and reticulocytes. Then, in optional step 206, one or more features and metrics are determined, such as, for example, reticulocyte hemoglobin measurements. For purposes of this disclosure, the determination of reticulocyte characteristics and metrics is shown as a separate step 206. However, in some embodiments, certain reticulocyte characteristics and metrics, such as area, perimeter, and shape, are measured as part of step 202 (e.g., identifying a representative set of red blood cells). This step 202 of identifying a representative set of red blood cells is now described in more detail.

Using the image obtained via the detector 106, the intensity value of each pixel in the sample image can be correlated with optical density values used in selecting a representative set of cells, and subsequently in artifact rejection, feature measurement, and mesh cell identification. The transmitted light intensity T (x, y) at a given image pixel (x, y) is related to the absorption coefficient α and path length (x, y) of incident light through the portion of the sample corresponding to that pixel:

T(x,y)=10^-α·(x,y)（6）

for each pixel in the image, the ratio of the pixel intensity to the maximum possible pixel intensity (e.g., pixel intensity/255 at 8-bit resolution) represents the proportion of light transmitted over the spatial location of the pixel. The proportion of transmitted light can be expressed in Optical Density (OD) units by taking the logarithm of equation (6):

OD(x,y)=-log(T)=α·(x,y)（7）

this process may be repeated for each pixel in the sample image. Thus, the optical density at each pixel in each image corresponds to the total amount of absorbing material (e.g., the product of the absorption coefficient and the thickness) in the sample at the location corresponding to the pixel.

Fig. 4 shows a flow chart 420 comprising a series of steps for selecting a representative set of red blood cells from an image of a prepared sample of blood. After acquiring the image of the sample, the electronic control system 108, and in particular the electronic processor 114, processes the image information to distinguish the cells included in the set of representative red blood cells from other cell types (e.g., white blood cells and platelets), cell clumps, and artifacts present in the sample.

First, in step 422 of fig. 4, the system 100 locates red blood cells (e.g., mature red blood cells and reticulocytes) in one or more sample images for further processing. Red blood cells typically absorb blue light (e.g., 415 nm) due to the presence of hemoglobin in the cell. However, white blood cells do not contain hemoglobin and therefore do not absorb blue light in the same manner as red blood cells. Images of the sample obtained under blue light can be used to identify red blood cells; in such images, white blood cells appear lighter and distorted because these cells minimally absorb blue light, thereby reducing their contribution to the image, often making them unrecognizable.

In some embodiments, a thresholding step may be used to ensure that the system 100 only identifies red blood cells for further analysis. For example, the system 100 may utilize only image pixels with intensity (or grayscale) values below 160 (for images captured at 8-bit resolution). Other intensity value thresholds ranging from 100 to 180 may be used to identify red blood cells from the image, while excluding white blood cells from further consideration.

Next, in step 424, the system 100 identifies a set of pixels for each red blood cell in the sample image. A variety of different methods can be used to identify the set of pixels associated with the cell. For example, in some embodiments, the system 100 performs the identifying step using a connected component tagging process. This process associates individual pixels from the sample image with objects in the image. For example, any two pixels separated by a pixel not designated as background are assigned to the same cell.

Additionally, the system 100 may exclude pixels located in boundary regions of the cell from making certain measurements related to, for example, cell volume or composition analysis. In particular, red blood cells tend to have thick and dark boundaries due to the way these cells refract the illumination light. Due to this refraction, the optical density of these pixels is often unreliable. After the connected component labeling process is completed, the system 100 may apply pixel blur layer erosion to the identified cells to remove the outermost n layers of pixels (e.g., pixels corresponding to the most refracting boundary regions). In general, depending on the magnification of the image, the pixel blur layer erosion may be selected to remove any number n of pixel layers (e.g., one pixel layer or more, two pixel layers or more, three pixel layers or more, four pixel layers or more, five pixel layers or more, six pixel layers or more, eight pixel layers or more, ten pixel layers or more). For example, it has been experimentally determined that erosion of the pixel blur layer comprising the outermost 0.5 μm around the red blood cells is generally suitable to significantly reduce the erroneous contribution to the measurements of cell volume and hemoglobin content for red blood cells each pixel corresponding to a fraction of the cell equal to 0.148 μm x 0.148 μm. Using the pixel sets corrected for by the blur layer erosion, various cell characteristics can be measured, such as average and maximum optical densities per cell that contribute to cell volume and composition analysis.

In step 426, the system 100 continues to identify a representative set of red blood cells from the sample image by confirming that the set contains only whole and normal shapes and sizes of red blood cells. In general, step 426 functions to discard the defective cells, overlapping cells, cell clusters, platelets, and non-cellular artifacts from inclusions in the set of representative red blood cells. For example, cells that are cut by or contacted with the edges of the image frames may be excluded from further analysis, thereby preventing inaccurate measurements. In addition, malformed cells that may exhibit changes in their non-standard shape in a defined cell volume may be excluded from the analysis. Also, measurements obtained from overlapping cells that may not be reliable for calculating cell volume or constituent content may be excluded from the set of representative cells. For these reasons, the shape of each identified cell is examined in step 426 and malformed and/or overlapping cells are excluded from further analysis.

A variety of different methods can be used to examine the shape of the identified cells. For example, in some embodiments, the shape of each cell can be examined by comparing the perimeter and area of the cell. Fig. 5 shows a schematic diagram of such a comparison. In fig. 5, a cell 500 is identified as a set of pixels in a sample image. Pixels corresponding to the boundaries of the cell 500 are shaded lighter than the interior pixels in fig. 5 for exemplary purposes, but they do not necessarily appear to be so in an actual image. The area of the cell 500 may be determined by counting the number of pixels in the group.

The pixel perimeter is determined from the boundary pixels using the set of pixels corresponding to the cell 500. This can be done by connecting lines through the center of each surrounding pixel to form a polygon in the image, and measuring the perimeter of the polygon. The ratio of the square of this cell perimeter value to the cell area value (i.e., the area of the polygon) is determined to examine the shape of the cell. For an ideal, perfectly circular cell, the value of this ratio is 4 π. The value of this ratio increases as the cell shape deviates from a circular contour. Using this criterion, cells with a ratio of the square of the perimeter to the area exceeding a minimum of 4 pi to a threshold value or greater were excluded from further analysis. Typically, the threshold is a few percent (e.g., 5% or more, 10% or more, 15% or more, 20% or more, 25% or more) of the minimum value of 4 π.

In addition to excluding individual malformed cells for further analysis, the process discussed above may also exclude overlapping cells. In the sample image, overlapping cells typically appear as larger, individual malformed cells (the change in transmitted light intensity is caused by an increase in the thickness of the material through which the incident light passes). When applying an analysis algorithm to such an image, overlapping cells are typically identified as a single larger cell. Thus, when a comparison of cell perimeter and area is made, the ratio falls well outside the threshold of allowable deviation from the ideal and overlapping cells are excluded.

Another method of examining the shape of the identified cell utilizes a convex hull represented by a polygon of the cell outline described above, and compares the area enclosed by the convex hull with the cell area determined from the image pixels. A high ratio of convex hull area to cell area can be used to identify irregularly shaped cells and exclude such cells from further analysis. Fig. 6 is a schematic diagram including two cells 600A and 600B. The contour of cells 600A and 600B is labeled 602A and 602B, respectively, in FIG. 6. A convex hull 604A is drawn around cell 600A and a convex hull 604B is drawn around cell 600B. As shown in fig. 6, the difference between the convex hull area and the cell area is greater for cell 600A than for cell 600B. Given the high degree of irregularity of the cell 600A, the cell 600A may be excluded from the set of representative red blood cells.

In some embodiments, the cell area measurement may be used in step 426 to exclude artifacts and overlapping cells from the set of representative red blood cells. For example, for red blood cell measurements that include reticulocyte recognition, only cells having an area ranging from 35 square microns to 65 square microns may be considered. Imaged objects with an area of less than 35 square microns are typically not red blood cells, but rather are artifacts like a piece of dust in the sample. Similarly, imaged objects with an area greater than 65 square microns are also typically not red blood cells; such an object may correspond to a drop of stain or several overlapping cells. Although the previous example describes a range of 36 to 65 square microns, other ranges may be used to select red blood cells to measure (e.g., 20 square microns to 80 square microns). And the range can be scaled according to the average cell size in the sample, thereby accounting for variability from patient to patient. It has been determined experimentally that although the 35 to 65 square micron range may exclude some red blood cells, such a range is more effective in eliminating artifacts from the sample image than the 20 to 80 square micron range.

The optical density value may be used to select the set of representative red blood cells in the sample. For example, if the average optical density value of an object imaged under blue light is too low, the object may be a white blood cell nucleus rather than a red blood cell. A mean optical density threshold (e.g., mean optical density of 0.33 or less) may be used for images acquired using blue light, with white blood cells excluded from the set of representative red blood cells of the sample (e.g., cells with mean optical density of 0.33 or less are likely to be white blood cells). For images obtained under blue or yellow illumination, an average optical density value of the object exceeding a certain threshold (e.g., an average optical density value of 0.66 or greater) may be used to identify stacked, overlapping, and/or clumped red blood cells that may be excluded from further analysis (e.g., red blood cells with an average optical density value of 0.66 or greater are likely to overlap another red blood cell).

The process shown in fig. 4 is terminated in step 428 with the final determination of a representative set of cells for further analysis. The measured characteristics of the cells within the representative set may then be used for reticulocyte identification, as described above.

(iii) Determining reticulocyte metrics and characteristics

Returning to fig. 2, after the representative set of red blood cells is identified in step 202, and reticulocytes are located (and optionally counted) among the red blood cells in step 204, one or more reticulocyte metrics and characteristics can optionally be determined in step 206. As discussed above, some reticulocyte characteristics including reticulocyte area, shape, perimeter, size, and optical density may be determined during step 202 while identifying the set of representative red blood cells. These characteristics may also be determined in step 206. In addition, other characteristics and metrics associated with the reticulocytes can be determined in step 206. For example, the systems and methods disclosed herein may use a combination of reticulocyte image features to calculate a reticulocyte metric such as volume and a component value such as hemoglobin content. Such combinations typically include, but are not limited to, combinations of such reticulocyte characteristics that the inventors have found to produce accurate, reproducible results for a wide variety of samples.

Once reticulocytes are located among the set of representative red blood cells as described above, some or all of the features disclosed herein can be calculated for each reticulocyte based on one or more images of the cells obtained by the system 100. A first set of features that may be calculated for the reticulocytes is a color-specific integrated optical density iod (c) that may be determined as follows:

IOD(c)=A·OD_mean(c)（8）

wherein A is the area of reticulocytes, and OD_mean(c) Is the average optical density in the image of the reticulocytes when the reticulocytes are illuminated with light of color c. If images of reticulocytes are acquired at different illumination wavelengths, the value of IOD (c) can be calculated for reticulocytes at each illumination wavelength. Fig. 7 shows a schematic image obtained with illumination of light of color c of a representative reticulocyte 700 identified by the process described in connection with flowchart 420. The image of the reticulocyte 700 includes a plurality of pixels. Average optical density OD of pixels in reticulocyte 700_mean(c) Corresponding to the sum of the pixel intensities in fig. 7 divided by the number of pixels in the image.

A second set of characteristics that can be calculated for each reticulocyte is the specific color volume vol (c) of the reticulocyte. The volume of the reticulocyte 700 in fig. 7 is calculated by summing the optical density value for each pixel corresponding to the reticulocyte 700. First, the "height" of the reticulocyte 700 on each pixel can be estimated as follows:

wherein OD_pixelIs the optical density associated with a given pixel, and OD_maxIs the maximum optical density among all the optical densities associated with the reticulocyte pixels. Thus, for example, pixel 720 in the image of reticulocyte 700 has an optical density that is less than the maximum optical density associated with pixel 710. The contribution of pixel 720 to the volume of reticulocyte 700 is the ratio OD₇₂₀/OD_maxWherein OD₇₂₀Is the optical density, OD, of the pixel 720_maxIs the optical density of the pixel 710. Then, the color-specific reticulocyte volume V is calculated by summing the ratio of the pixel optical density to the maximum optical density over all pixels in the reticulocyte 700:

（10）

wherein the number N of pixels in the reticulocyte 700 is used in equation (10)_pixelsAnd the average pixel optical density OD of the pixels in the reticulocyte 700_meanEach of the replacement and reticulocytes 700The sum of the optical densities associated with the pixels.

In general, the optical density values of pixels near the edge of the reticulocyte are not an effective contributor to the specific color volume measurement because the light refracted on the edge of the cell creates an artificial dark edge around the reticulocyte. To avoid this effect from such edge pixels, the system may erode the blur layer on the edges of the reticulocyte by one or more pixels as previously described, measure the average and maximum optical densities of the blurry regions of the reticulocyte, and thereafter extrapolate to the edges of the reticulocyte by multiplying by the area of the entire unetched blur layer.

Further, when a plurality of images corresponding to different illumination wavelengths are used to acquire an image of a single reticulocyte, a reticulocyte volume calculation determination may be made at each color of illumination light. Thus, the specific color reticulocyte volume can be determined as follows:

wherein A is the area of the whole reticulocyte surrounding the cell, OD_mean(c) Is the color-specific mean optical density, and OD, of the pixels within the obscured region of the reticulocyte_max(c) Is the maximum optical density of a particular color of the erosion cloud of reticulocytes (e.g., a graphPixel 710 in 7). The calculated color-specific reticulocyte volume, vol (c), may be scaled to express reticulocyte volume in appropriate units (e.g., femtoliters).

In some embodiments, it may be useful to add one or more correction factors to equation (11) to adjust for the event that some darkness in the reticulocyte image may not actually be caused by the RNA or hemoglobin content of the reticulocytes. In addition, a scaling factor may be applied to convert the volume measurement value into a unit of measurement such as fly-lift (fL). To account for these correction and scaling factors, equation (11) may be rewritten as:

where S corresponds to a scaling factor or slope, J corresponds to a correction factor that accounts for the deviation in determining the maximum optical density, and B corresponds to an intercept value corresponding to a global offset value.

The correction factor, scaling factor, and intercept value may be determined experimentally by using a data set of known volume values for a plurality of blood samples processed on, for example, a calibration flow cytometer. A slightly different set of correction factors will generally provide the best results for each different sample, but the correction factors can be determined from the results across the entire data set. For example, for a data set containing known reticulocyte volumes of 1,000 blood samples, the correction factor that works best across the entire data set can be determined by selecting a correction factor that minimizes the sum of squared differences between measured and expected volume values across the entire data set. The scaling factor may be determined across the entire data set by selecting a scaling factor that best converts the raw volume values into the desired units of measurement like fly-up. An intercept value B may be calculated for the data set to ensure that equation (22) passes through the origin when the data is presented on a two-dimensional graph. The correction factor, scaling factor, and intercept value may be stored in a memory unit associated with the electronic control system 108, retrieved from memory when determining the specific color reticulocyte volume as shown in equation (12) for analyzing a new sample.

When equation (8) and equation (11) or (12) are used, two features (e.g., integrated optical density iod (c) and volume vol (c)) may be determined for each color of illumination light used to acquire an image of the sample. For example, if four different colors of illumination light are used, a total of eight different characteristics may be determined for each reticulocyte identified from the representative set. In addition, as described above, the area a of each individual reticulocyte may be determined from the image of the reticulocyte. The specific color combined optical density and reticulocyte volume, and reticulocyte area can then be used to calculate various metrics for each reticulocyte.

Reticulocyte metrics such as reticulocyte volume and constituent number can be calculated based on a weighted combination of some or all of the features disclosed above as being calculated for reticulocytes. In general, the metric M can be determined according to the following equation:

where n corresponds to each color of illumination light, ω, used to acquire an image of the reticulocytes_n,iThe value is the specific color weight coefficient, ω, of the integrated optical density IOD (n) of each specific color_n,vThe value is the specific color weight coefficient, ω, for each specific color volume Vol (n)_nIs the weighting factor for reticulocyte area A, and K is the offset value. For example, when four different illumination wavelengths are used to acquire an image of a reticulocyte (e.g., red = r, yellow = y, red = g, and blue = b), then the reticulocyte volume V may be determined as:

V=ω_r,i·IOD(r)+ω_y,i·IOD(y)+ω_g,i·IOD(g)+ωb_,i·IOD(b)

+ω_r,v·Vol(r)+ω_y,v·Vol(y)+ω_g,v·Vol(g)+ω_b,v·Vol(b)

+ω_a·A+K

（14）

the number of reticulocyte components can be determined in a similar manner. For example, the concentration H of hemoglobin in reticulocytes can be calculated as follows:

H=ω_r,i·IOD(r)+ω_y,i·IOD(y)+ω_g,i·IOD(g)+ω_b,i·IOD(b)

+ω_r,v·Vol(r)+ω_y,v·Vol(y)+ω_g,v·Vol(g)+ω_b,v·Vol(b)

+ω_a·A+K

（15）

the difference between equations (14) and (15) above is in the weighting factor and the value of the offset K. When equations (14) and (15) are used, the reticulocyte volume and the number of components (e.g., the number of hemoglobins) can be determined for a plurality of reticulocytes in the sample. The results may be averaged to determine the average reticulocyte volume and the average concentration of constituents of the sample (e.g., average reticulocyte hemoglobin).

The weighting coefficients associated with a particular color feature in equation (13) may be determined from available training data, for example, by determining linear regression coefficients that map experimentally determined sample features to training data containing known volume and/or constituent concentration values for such sample. Determining specific color weights using linear regression practices can improve the accuracy of the sample average reticulocyte volume and average reticulocyte component concentration measurements by correcting uncontrollable factors that affect the measurements, such as membrane thickness and cell-to-cell variability in stain absorption. After determining a particular color weight value from the training data, the weight value may be stored and later retrieved from a storage unit (e.g., an electronic storage unit) prior to analyzing each sample.

In general, a wide variety of different samples can be used to determine the appropriate weighting coefficients. In order to obtain highly reproducible results, it may be advantageous to use training data that spans the entire range of values of the calculated parameters. Also, if the sample to be analyzed includes unusual morphological features like cell clumps, it may be advantageous to use training data for representative samples that include such features.

As an example, after determining a set of weighting coefficients from a set of training data for determining reticulocyte volume, equation (14) may be rewritten as follows:

V=(-4.04)·IOD(r)+8.49·IOD(y)+(-3.69)·IOD(g)+4.40·IOD(b)

+4.68·Vol(r)+(-8.20)·Vol(y)+3.57·Vol(g)+0.0159·Vol(b)

+(-0.125)A+4.84（16）

similarly, after determining a set of appropriate weighting coefficients from a set of training data for determining reticulocyte hemoglobin, equation (15) can be rewritten as follows:

H=(-1.05)·IOD(r)+(-2.44)·IOD(y)+1.12·IOD(g)+2.15·IOD(b)

+1.95·Vol(r)+(-0.112)·Vol(y)+(-1.27)·Vol(g)+0.457·Vol(b)

+(-0.221)·A+(-5.73)（17）

additional details regarding techniques for measuring cell volume and cellular component content are disclosed in the following references: co-pending U.S. patent application nos. 13/446,967 and 13/447,045, filed 4/13/2012, and incorporated herein by reference in their entirety.

As previously mentioned, the systems and methods disclosed herein can be used to analyze both a white blood cell sample (e.g., a sample taken from a patient) and a quality control composition. The weighting coefficients shown in equations (14) - (17) can be used to analyze both the white blood cell sample and the quality control complex. In other embodiments, a weighting factor specific to the quality control compound may be used; as described previously in connection with patient blood samples, such coefficients may be determined using training data comprising known measurements of reticulocytes in the quality control complex.

In some embodiments, the device may be used to analyze the control complex to assess the accuracy of the results produced by the device. For example, the results of the device analyzing the control complex (e.g., determining the number of reticulocytes and the like reticulocyte hemoglobin, reticulocyte volume, mean reticulocyte hemoglobin, and mean reticulocyte volume) can be compared to reference values for these numbers of control complex to assess the accuracy of the device. If the difference between these number of one or more of the determined values and the reference value exceeds a threshold value, the device may be recalibrated. Recalibration may include, for example, re-determining the values of some or all of the weighting coefficients in equations (14) - (17) from a reference blood sample, as described herein.

Other embodiments

In equations (14) - (17), the illumination light of four colors (red, yellow, green, and blue) is used to illuminate the sample, and the integrated optical density and reticulocyte volume are calculated from the images corresponding to each of these colors. The illumination wavelengths used to calculate the specific color integrated optical density values and volumes may be, for example, 635nm, 598nm, 525nm, and 415nm, although other values within the red, yellow, green, and blue regions of the electromagnetic spectrum may be used in other embodiments. More generally, a different number of illumination wavelengths may be used, and images corresponding to each illumination wavelength may be acquired and used to calculate a particular color value and/or reticulocyte volume for the integrated optical density. For example, in some embodiments, three different wavelengths of light are used to illuminate the sample, in some embodiments, more than four wavelengths of illumination light (e.g., five wavelengths, six wavelengths, seven wavelengths, eight wavelengths, ten wavelengths) may be used, and a particular color integrated optical density, cell volume, and weight coefficient may be determined at some or all of the illumination wavelengths. In general, the wavelength of the illumination light may be selected such that each image at each different wavelength includes different information about the sample. For example, if the sample includes three spectral contributors, the three wavelengths of illuminating light may be selected for use so that each of the three wavelengths is most strongly absorbed by a different one of the spectral contributors.

As discussed above, the specific color weight coefficients in equation (13) may be determined by mapping linear regression coefficients of experimentally determined features of a large number (e.g., 1,000) of blood samples to a training data set containing known values of reticulocyte volume and/or concentrations of various reticulocyte components of such samples obtained from, for example, a calibrated flow cytometer system. As sample preparation parameters change (e.g., modification of stain complexes that affect the appearance of stained reticulocytes or other factors that affect how the reticulocytes absorb the stain, such as the degree to which the sample is dried prior to fixation and staining), the process of determining the specific color weights and intercept values of equation (13) can be repeated to ensure that accurate volume and reticulocyte component measurements are determined for a given set of sample preparation parameters. However, once the sample preparation parameters have been optimized for a particular sample preparation system, the experimentally derived weight coefficients and other parameter values in equation (13) will result in accurate and reproducible measurements of reticulocyte volume and/or the number of reticulocyte components.

In equation (13), the metric M is calculated as a weighted linear combination of the color-specific integrated optical density, the color-specific volume, and the reticulocyte area. However, not all of these features are used in all embodiments to determine the value of the metric. For example, in some embodiments, the metric M may be calculated as a weighted combination of only specific color integrated optical densities or specific color volumes. In certain embodiments, the metric M may be calculated as a weighted combination of reticulocyte area and color-specific integrated optical density or color-specific volume. In some embodiments, the metric M may be calculated as a weighted linear combination of the color-specific integrated optical density and the color-specific volume. In general, the appropriate combination of features for calculating a particular metric M may be determined using a reference sample of known values of metric M.

When determining the amount of a particular component in a reticulocyte, if only a single spectral contributor (e.g., an absorption contributor such as hemoglobin) is present in the sample, the total amount of that contributor present in the particular reticulocyte may be determined by summing the intensity contributions from each pixel in the image corresponding to the selected reticulocyte. Since the intensity contributors correspond only to hemoglobin absorption, only one sample image is needed to determine the total amount of hemoglobin present in the reticulocytes.

In practice, however, the sample is typically prepared with one or more stains to assist a technician or automated imaging system in identifying, counting, and classifying the various cell types. For a plurality of spectral contributors in the sample, the absorption at each illumination wavelength is a combination of the absorption caused by each contributor in the sample; the total contribution of a particular reticulocyte at any wavelength still corresponds to the sum of the contributions from each pixel representing the reticulocyte at that wavelength. Thus, for the spectral contributors present in the sample-hemoglobin (H), eosin (E), and sky blue (a), and assuming that the three images of the sample correspond to illumination light with central wavelengths in the yellow (y), green (g), and blue (b) regions of the electromagnetic spectrum, it can be assumed that the optical density of a particular reticulocyte (or, all pixels in the image corresponding to one or more reticulocytes) at each of these three wavelengths is a linear combination of the absorption at each wavelength induced by each spectral component:

OD(y)=H·α_y,H+E·α_y,E+A·α_y,A

OD(g)=H·α_g,H+E·α_g,E+A·α_g,A（18）

OD(b)=H·α_b,H+E·α_b,E+A·α_b,A

α therein_i,jIs the absorption coefficient of a particular contributor j (e.g., hemoglobin H, eosin E, or sky blue a) at a wavelength i (e.g., yellow y, green g, or blue b).

Each center wavelength of light can be determined by passing a known spectrum through the sample to a detector and measuring the absorbance of the sample. For example, the detector may acquire three images of the sample with illumination sources having narrow illumination spectra in the yellow, green, and blue regions, respectively. In case each spectral contributor has an absorption spectrum containing a local maximum, the illumination source may be selected such that the emission spectrum corresponds to or best approximates the local maximum of the spectral contributor. For example, blue light illumination may be selected to a wavelength (e.g., 415 nm) corresponding to the peak absorbance of hemoglobin in the sample. The yellow light illumination may be associated with a wavelength (e.g., 598 nm) corresponding to a peak absorbance of a sky-blue stain in the sample. Similarly, the green light illumination may be selected to be a wavelength (e.g., 525 nm) corresponding to the peak absorbance of an eosin stain in the sample. Additional illumination wavelengths may be selected to correlate with peak absorbance values of additional spectral contributors in the sample.

The optical density quantities OD (y), OD (g) and OD (b) may be determined from the image information, and the absorption coefficient α may be obtained from a reference source or determined experimentally_i,j. Thus, the system of equations (18) includes three unknowns-H, E and a, which can be solved to derive the number of each of the three spectral contributors present in each reticulocyte, or collectively for all cells including all reticulocytes in the sample if the pixels selected for analysis collectively correspond to all identified cells in the image.

However, the methods and systems disclosed herein provide a simpler, more efficient method of determining the number of reticulocyte components. As described above, equation (13) with appropriate weighting coefficients can be used to determine the amount of constituents for only those constituents of interest, increasing the speed at which sample analysis can be completed. Also, in complex samples where the number of spectral contributors is not well understood, it may be difficult to construct a system of equations like in equation (18). However, equation (13) allows the determination of the amount of a particular reticulocyte component even if the presence of other spectral contributors in the reticulocyte is not well determined. Thus, while in some embodiments the spectral contributions from hemoglobin, eosin, and sky blue in equation set (18) may be distinguished by acquiring images at three different illumination wavelengths, determining the numerical values of reticulocyte metrics such as reticulocyte hemoglobin, reticulocyte volume, mean reticulocyte hemoglobin, and mean reticulocyte volume using more than three features and/or more than three illumination wavelengths as described herein allows for the correction of other systematic and non-systematic sources of error when measuring a blood sample.

In addition to specific color integrated optical density and volume, and reticulocyte area, other image features may be used to determine reticulocyte volume and/or number of reticulocyte components. For example, in some embodiments, equation (13) may include an additional term corresponding to the product of the reticulocyte perimeter and the weight coefficient. As described above, an appropriate weighting factor may be determined for the reticulocyte perimeter from the training data. More generally, a variety of additional terms derived from the reticulocyte image that determine appropriate weighting coefficients from the training data may be included in equation (13). Such terms may include image geometry related to morphology of reticulocytes and/or specific color measurements of integrated optical density and volume at more than three or four illumination wavelengths. Without wishing to be bound by theory, the additional term may allow simultaneous fitting of the reference sample information to determine the number of all weighting factors to correct for effects such as imaging aberrations, absorption from other sample components, and systematic measurement errors that are not fully accounted for by the model of equation (13). For example, it has been found that the inclusion of the integrated optical density and reticulocyte volume terms corresponding to the wavelength of red light illumination and the term corresponding to the reticulocyte area improves the accuracy of determining reticulocyte hemoglobin in many samples as compared to measurement techniques that do not use sample images or reticulocyte area measurements obtained at the wavelength of red light illumination.

In general, the methods and systems disclosed herein can be used to determine the amount of naturally occurring components in a sample (e.g., hemoglobin in reticulocytes) and/or the amount of components added to a sample (e.g., stain applied and bound to reticulocytes). Also, in some embodiments, the methods and systems disclosed herein may be used to determine the amount of more than one constituent present in a sample. For example, the amount of two or more components may be determined by applying an appropriate stain and/or selecting a center wavelength appropriate for the sample image. Consider a sample that includes hemoglobin as a naturally occurring absorbing component. The sample may be stained with two broad absorption stains S (1) and S (1), and with a third stain S (3) having a relatively narrow absorption band. S (3) selectively binds to a particular component of interest in the reticulocytes, such that measuring the amount of S (3) present results in a measurement of the component.

If the absorption spectra of hemoglobin and S (3) are sufficiently spectrally separated that hemoglobin is only at wavelength λ₁，λ₂And λ₃Above rather than at λ₄Has significant absorption, and S (3) is only at the wavelength lambda₂，λ₃And λ₄Above rather than at λ₁Then, assuming that S (1) and S (2) have significant absorption at all four wavelengths, the number of reticulocyte components can be measured from the irradiation wavelength λ according to the method for measuring reticulocyte component number disclosed herein₁，λ₂And λ₃Determining the amount of reticulocyte hemoglobin in the corresponding image of the sample, and according to the same method, from the wavelength λ of the radiation₂，λ₃And λ₄The amount of S (3) is determined in the image of the corresponding sample. This approach can be further generalized to a greater number of components of interest, and a greater or lesser number of broad absorption spectrum contributors like S (1) and S (2).

Reticulocyte result reporting

In certain embodiments, display 110 may be used, for example, to display the identified reticulocytes, the determined reticulocyte volume, the number of components, the average reticulocyte volume, and/or the average component concentration to a system operator. The volume and composition results may be displayed as reticulocytes, or as an average result across the sample. Further, the calculated numerical results (e.g., for each reticulocyte) may be superimposed over one or more images of the reticulocyte. In general, the system operator may exercise control over the manner in which results are displayed using the human interface unit 112 (e.g., keyboard, mouse, and/or other input devices). The system operator may also exercise control of any other parameters, conditions, and options associated with the methods disclosed herein via the interface unit 112 and display 110.

One or more metrics may also be calculated and displayed from the average reticulocyte volume and/or the average reticulocyte hemoglobin measurement. For example, in some embodiments, the reticulocyte distribution width may be calculated and displayed for the operator. In addition, the average reticulocyte hemoglobin measurement value can be used along with the hematocrit value of the sample to calculate an average reticulocyte hemoglobin concentration. Additional details regarding methods and systems for displaying reticulocyte images and related reticulocyte metrics can be found in the following references: U.S. patent application No. 13/526,223, filed 6/18/2012, the contents of which are hereby incorporated by reference in their entirety.

The reticulocyte volume and constituent concentration measurements and/or metrics calculated therefrom may be stored with the sample image in an electronic memory unit associated with the control system 108. This information can be stored, for example, in an electronic medical record associated with the patient to which the sample 104 corresponds. Alternatively, or in addition, the information may be sent to one or more doctors or other treatment personnel. The information may be sent to the computing device via a network (e.g., a computer network). Also, the information may be sent to a handheld device, such as a mobile phone, which may include a warning or alarm if the metric is outside a predetermined range of values.

Methods and systems for red blood cell identification and measurement are disclosed, for example, in U.S. patent application No. 13/446,967 filed 4/13/2012, the contents of which are hereby incorporated by reference in their entirety.

Automatic sample preparation system

The systems and methods disclosed herein may be used on a variety of different automated sample preparation systems. Fig. 8 shows a schematic diagram of an embodiment of an automated sample preparation system 800. The system 800 includes a plurality of subsystems that store a substrate, deposit a sample on the substrate, examine the sample prepared on the substrate, and store the prepared sample.

The substrate storage subsystem 810 is configured to store substrates prior to deposition of samples thereon. The substrate may include, for example, a microscope slide, a coverslip, and similar flat, optically transparent substrates. The substrate may be formed of a variety of different amorphous or crystalline materials, including various types of glass. Subsystem 810 may include a manipulator that selects individual substrates from a storage container and transfers the selected substrates to a sample deposition subsystem 820.

The sample deposition subsystem 820 deposits a selected amount of a sample of interest, such as a blood sample, on a substrate. Generally, subsystem 820 includes various fluid transfer components (e.g., liquid pumps, fluid tubes, valves) configured to deposit a sample. The fluid transfer assembly may also be configured to expose the substrate to various types of solutions, including wash solutions, one or more staining solutions combined with the sample, fixative solutions, and/or buffer solutions. Subsystem 820 may also feature a fluid removal assembly (e.g., a vacuum subsystem) and drying apparatus to ensure that the sample is immobilized on the substrate. The substrate handler may transfer the sample-bearing substrate to the inspection subsystem 830.

Inspection subsystem 830 includes various components that acquire images of samples on a substrate and analyze the images to determine information about the samples. For example, inspection subsystem 830 may include one or more light sources (e.g., light emitting diodes, laser diodes, and/or lasers) that direct incident light toward the sample. Imaging subsystem 830 can also include optics (e.g., a microscope objective) that capture transmitted and/or reflected light from the sample. A detector (e.g., a CCD detector) coupled to the optics may be configured to capture an image of the sample. Information derived from analyzing the image of the sample may be stored on a variety of optical and/or electrical storage media for later retrieval and/or further analysis.

After inspection, the substrate handler may transfer the substrate to the storage subsystem 840. Storage subsystem 840 may mark, for example, information about the source of the sample applied to the substrate, the time of analysis, and/or any irregularities identified during analysis on the various substrates. The storage subsystem 840 may also store processed substrates in multi-substrate holders that may be removed from the system 800 when filled with substrates.

As shown in fig. 8, each of the various subsystems of system 800 may be linked to a common electronic processor 850. The processor 850 may be configured to control the operation of each subsystem of the system 800 in an automated manner with relatively little (or no) input from a system operator. Results from the sample analysis may be displayed on the system display 860 for use by a system management operator. An interface 870 allows an operator to issue commands to the system 800 and manually review the results of the automated analysis.

Additional aspects and features of automated sample processing systems are disclosed, for example, in the following patent documents: U.S. patent application No. 13/293,050, filed on 9/11/2011, is hereby incorporated by reference in its entirety.

Hardware and software implementation

The method steps and processes described herein may be implemented in hardware or software, or a combination of both. In particular, the electronic processor 114 may include software and/or hardware instructions to perform any of the methods disclosed above. The method may be implemented in a computer program using standard programming techniques, following the method steps and figures disclosed herein. Program code is applied to input data to perform the functions described herein. Applying the output information to one or more output devices, such as a printer or a display device, or accessing a website, such as a web page on a computer monitor for remote monitoring.

Each program is preferably implemented in a high level procedural or object oriented programming language to communicate with a processor. However, the programs may be implemented in assembly or machine language, if desired. In any case, the language may be a compiled or interpreted language. Each computer program may be stored on a storage medium or device (e.g., electronic memory) readable by a processor to configure and operate the processor to perform the processes described herein.

FIG. 9 is a schematic diagram of a computer system 900 that may be used to control the operations described in connection with any of the computer-implemented methods described herein, according to one embodiment. The system 900 includes a processor 910, a memory 920, a storage device 930, and an input/output device 940. Each of the components 910, 920, 930, and 940 are interconnected using a system bus 950. The processor 910 is capable of processing instructions for execution within the system 900. In some embodiments, the processor 910 is a single-threaded processor. In another embodiment, the processor 910 is a multi-threaded processor. The processor 910 is capable of processing instructions stored in the memory 920 or on the storage device 930 to display graphical information for a user interface on the input/output device 940. The processor 910 may be substantially similar to the processor 850 described above with reference to fig. 8.

Memory 920 stores information within system 900. In some embodiments, memory 920 is a computer-readable medium. The memory 920 may include volatile memory and/or nonvolatile memory.

The storage device 930 is capable of providing a mass storage for the system 900. In general, storage device 930 may include any non-transitory tangible medium configured to store computer-readable instructions. In one embodiment, storage device 930 is a computer-readable medium. In various different embodiments, storage device 930 may be a floppy disk device, a hard disk device, an optical disk device, or a tape device.

Input/output device 940 provides input/output operations for system 900. In some embodiments, input/output devices 940 include a keyboard and/or pointing device. In some embodiments, input/output device 940 includes a display unit that displays a graphical user interface. In some embodiments, input/output devices 940 include one or more of the displays 860 and interfaces 870 described above with reference to fig. 8.

The features may be implemented in digital electronic circuitry, or in computer hardware, firmware, or in combinations of them. The features can be implemented in a computer program product tangibly embodied in an information carrier, e.g., in a machine-readable storage device, for execution by a programmable processor; and such features may be implemented by a programmable processor executing a program of instructions to perform functions of the described embodiments by operating on input data and generating output. The described features can be implemented in one or more computer programs that are executable on a programmable system including at least one programmable processor coupled to receive data and instructions from, and to transmit data and instructions to, a data storage system, at least one input device, and at least one output device. A computer program comprises a set of instructions that can be used, directly or indirectly, in a computer to perform a certain activity or bring about a certain result. A computer program can be written in any form of programming language, including compiled or interpreted languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

Various software architectures may be used to implement the methods and systems described in this application. For example, a publish/subscribe messaging schema may be used in implementing the methods and systems described herein. In the case of publish/subscribe messaging, the system includes several hardware and software modules that communicate only via the messaging module. Each module may be configured to perform a particular function. For example, the system may include one or more of a hardware module, a camera module, and a focusing module. The hardware module may send commands to imaging hardware that implement fast auto-focus, which in turn triggers the camera to acquire images. In some embodiments, the hardware module may include the control system 108 described above with reference to fig. 1.

The camera module may receive images from the camera and determine parameters such as shutter time or focus. The images may also be buffered in the memory of the computer before being processed by the camera module. The camera module may also send a message to interrupt the hardware module when an image is seen that is sufficient to determine the appropriate shutter time or focus when an initial search is made for the tilt of the slide. In some embodiments, the camera module includes the detector 106 described above with reference to fig. 1.

The system may also include a focusing module, which may be implemented in software, hardware, or a combination of software and hardware. In some embodiments, the focus module examines all frames in a stack and estimates how far the stack is from an ideal or ideal focal distance. The focus module may also be responsible for assigning a focus score to each frame in a stack of images.

Processors suitable for the execution of a program of instructions include, by way of example, both general and special purpose microprocessors, and the sole processor or one of multiple processors of any kind of computer. Generally, a processor will receive instructions and data from a read-only memory or a random access memory or both. A computer includes a processor that executes instructions and one or more memories that store instructions and data. Generally, a computer will also include, or be operatively coupled to communicate with, one or more mass storage devices for storing data files; such devices include magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and an optical disc. Storage devices suitable for tangibly embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices, such as EPROM, EEPROM, and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and memory may be supplemented by, or incorporated in, ASICs (application-specific integrated circuits).

To provide for interaction with a user, the features can be implemented on a computer having a display device such as a CRT (cathode ray tube) or LCD (liquid crystal display) monitor for displaying information to the user and a keyboard and a pointing device such as a mouse or a trackball by which the user can provide input to the computer. Alternatively, the computer may have no keyboard, mouse, or monitor attached thereto, and may be remotely controlled by another computer.

The features can be implemented in a computer system that includes a back-end component, such as a data processor, that includes a middleware component, such as an application server or an Internet server, that includes a front-end component, such as a client computer having a graphical user interface or an Internet browser, or any combination of them. The components of the system can be interconnected by any form or medium of digital data communication, such as a communication network. Examples of communication networks include, for example, a LAN (local area network), a WAN (wide area network), and computers and networks forming the internet.

The computer system may include clients and servers. A client and server are generally remote from each other, and typically interact through a network such as the one described. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

The processor 910 executes instructions related to a computer program. The processor 910 may include hardware such as logic gates, adders, multipliers and counters. The processor 910 may further include a separate Arithmetic Logic Unit (ALU) that performs arithmetic and logical operations.

OTHER EMBODIMENTS

It is to be understood that the foregoing description is intended to illustrate and not to limit the scope of the disclosure, which is defined by the scope of the appended claims. Other aspects, advantages, and modifications are within the scope of the claims. For example, while the foregoing description and the schematic in FIG. 1 discuss measuring transmitted light from a sample, the methods and systems disclosed herein may also be used when an image of a sample corresponds to light reflected from the sample. Some samples may be naturally reflective, or reflective markers may be added, so that the reflected light provides a convenient method of determining the composition and/or quantity of the cells. In some embodiments, sample 104 may be placed on top of a substrate, such as a microscope slide, with a reflective coating. The reflective coating may act to pass a transmitted light back through the sample a second time so that the measured "reflected" light actually corresponds to the incident light transmitted through the sample twice.

In general, the methods and systems disclosed herein can be used to determine reticulocyte count, volume per reticulocyte, constituent content, and/or average reticulocyte volume or average reticulocyte constituent of a sample for a variety of different sample types. For example, reticulocyte volume and components such as hemoglobin or other proteins may be determined for samples including reticulocytes from body fluids and tissues including blood, bone marrow, urine, epithelial tissues, tumors, semen, sputum, and other tissues or circulating or also circulating biological fluids.

Claims (42)

1. A method of identifying reticulocytes in a blood sample deposited on a substrate, the method comprising:

illuminating the sample with incident light at two different wavelengths, acquiring a two-dimensional image of the sample corresponding to a first wavelength, and acquiring a two-dimensional image of the sample corresponding to a second wavelength;

analyzing the image to identify a set of representative red blood cells;

determining the area of each red blood cell in the set;

determining a color value of each red blood cell in the set; and

for each red blood cell in the set, identifying the red blood cell as a reticulocyte if the area of the red blood cell exceeds an area cutoff value and the color value of the red blood cell is less than a color cutoff value,

wherein the color value of each of the red blood cells comprises a difference between optical density values of the red blood cells at two illumination wavelengths.

2. The method of claim 1, wherein one of the wavelengths is between 400nm and 457nm and the other wavelength is between 575nm and 600 nm.

3. The method of claim 1, wherein determining the color value of each of the identified red blood cells comprises, for each red blood cell:

determining a set of pixels associated with the cell;

determining, for each of the set of pixels, an optical density corresponding to the first wavelength;

determining, for each of the set of pixels, an optical density corresponding to the second wavelength;

determining, for the set of pixels, an average optical density corresponding to the first wavelength;

determining, for the set of pixels, an average optical density corresponding to the second wavelength; and

the difference between the average optical densities is calculated to determine the color value of the cell.

4. The method of claim 1, wherein determining the area of each red blood cell in the set comprises, for each red blood cell:

determining a set of pixels associated with the cell; and

the area of the cell is determined by counting the number of pixels in the group.

5. The method of claim 1, wherein determining the area of each red blood cell in the set comprises, for each red blood cell:

determining a set of pixels associated with the cell;

determining a polygon around the set of pixels; and

the area of the cell is determined by calculating the area of the polygon.

6. The method of claim 1, wherein the color cutoff value is based on a percentile of a distribution of color values of red blood cells.

7. The method of claim 6, wherein the percentile corresponds to 70% within a distribution of color values of red blood cells.

8. The method of claim 6, wherein the color cutoff value corresponds to a sum of a percentile and a color offset value.

9. The method of claim 8, further comprising determining percentile and color offset values based on a set of training data for which the number of reticulocytes is known.

10. The method of claim 1, wherein the area cutoff value is based on a percentile of an area distribution of red blood cells.

11. The method of claim 10, wherein the percentile corresponds to a 20% area distribution of red blood cells.

12. The method of claim 10, wherein the area cutoff value corresponds to a sum of the percentile and an area offset value.

13. The method of claim 12, further comprising determining percentile and area offset values based on a set of training data for which the number of reticulocytes is known.

14. The method of claim 1, further comprising excluding the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at one of the two wavelengths is greater than a cutoff value.

15. The method of claim 1, further comprising excluding the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at wavelengths other than the two wavelengths is greater than a cutoff value.

16. The method of claim 1, further comprising, for each red blood cell identified as a reticulocyte, determining a volume of the reticulocyte.

17. The method of claim 16, further comprising determining the volume of reticulocytes based on the integrated optical density of reticulocytes corresponding to the plurality of illumination wavelengths.

18. The method of claim 17, further comprising determining a mean reticulocyte volume parameter for the sample.

19. The method of claim 1, further comprising, for each red blood cell identified as a reticulocyte, determining a hemoglobin content of the reticulocyte.

20. The method of claim 19, further comprising determining a hemoglobin content of the reticulocytes based on a weighted combination of an area of the reticulocytes, a volume of the reticulocytes corresponding to the plurality of illumination wavelengths, and a combined optical density of the reticulocytes corresponding to the plurality of illumination wavelengths.

21. The method of claim 19, further comprising determining a mean reticulocyte hemoglobin value for the sample.

22. A system for identifying reticulocytes in a blood sample deposited on a substrate, the system comprising:

a light source configured to illuminate the sample with incident light at two different wavelengths;

a detector configured to acquire a two-dimensional image of the sample corresponding to a first wavelength and to acquire a two-dimensional image of the sample corresponding to a second wavelength; and

an electronic processor configured to:

analyzing the image to identify a set of representative red blood cells;

determining the area of each red blood cell in the set;

determining a color value of each red blood cell in the set; and

for each red blood cell in the set, identifying the red blood cell as a reticulocyte if the area of the red blood cell exceeds an area cutoff value and the color value of the red blood cell is less than a color cutoff value,

wherein the electronic processor is configured to determine a color value of each of the red blood cells based on a difference between optical density values of the red blood cells at the two illumination wavelengths.

23. The system of claim 22, wherein the first wavelength is between 400nm and 457nm and the second wavelength is between 575nm and 600 nm.

24. The system of claim 22, wherein for each red blood cell in the set, the electronic processor is configured to determine the color value of the cell by:

determining a set of pixels associated with the cell;

determining, for each of the set of pixels, an optical density corresponding to the first wavelength;

determining, for each of the set of pixels, an optical density corresponding to the second wavelength;

determining, for the set of pixels, an average optical density corresponding to the first wavelength;

determining, for the set of pixels, an average optical density corresponding to the second wavelength; and

the difference between the average optical densities is calculated to determine the color value of the cell.

25. The system of claim 22, wherein for each red blood cell in the set, the electronic processor is configured to determine the area of the cell by:

determining a set of pixels associated with the cell; and

the area of the cell is determined by counting the number of pixels in the group.

26. The system of claim 22, wherein for each red blood cell, the electronic processor is configured to determine the area of the cell by:

determining a set of pixels associated with the cell;

determining a polygon around the set of pixels; and

the area of the cell is determined by calculating the area of the polygon.

27. The system of claim 22, wherein the electronic processor is configured to determine the color cutoff value based on a percentile of a distribution of color values of the red blood cells.

28. The system of claim 27, wherein the percentile corresponds to 70% within a distribution of color values of red blood cells.

29. The system of claim 27, wherein the electronic processor is configured to determine the color cutoff value as a sum of a percentile and a color offset value.

30. The system of claim 29, wherein the electronic processor is configured to determine the percentile and the color offset value based on a set of training data in which the number of reticulocytes is known.

31. The system of claim 22, wherein the electronic processor is configured to determine an area cutoff value based on a percentile of an area distribution of the red blood cells.

32. The system of claim 31, wherein the percentile corresponds to a 20% area distribution of red blood cells.

33. The system of claim 31, wherein the electronic processor is configured to determine the area cutoff value as a sum of the percentile and an area offset value.

34. The system of claim 33, wherein the electronic processor is configured to determine the percentile and the area offset value based on a set of training data in which the number of reticulocytes is known.

35. The system of claim 22, wherein the electronic processor is configured to exclude the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at one of the two wavelengths is greater than a cutoff value.

36. The system of claim 22, wherein the electronic processor is configured to exclude the red blood cell from the representative set if a standard deviation of optical densities of pixels associated with the red blood cell at wavelengths different from the two wavelengths is greater than a cutoff value.

37. The system of claim 22, wherein for each red blood cell identified as a reticulocyte, the electronic processor is configured to determine a volume of the reticulocyte.

38. The system of claim 37, wherein the electronic processor is configured to determine the volume of reticulocytes based on the integrated optical density of the reticulocytes corresponding to the plurality of illumination wavelengths.

39. The system of claim 38, wherein the electronic processor is configured to determine an average reticulocyte volume parameter for the sample.

40. The system of claim 22, wherein for each red blood cell identified as a reticulocyte, the electronic processor is configured to determine a hemoglobin content of the reticulocyte.

41. The system of claim 40, wherein the electronic processor is configured to determine the hemoglobin content of the reticulocyte based on a weighted combination of an area of the reticulocyte, a volume of the reticulocyte corresponding to the plurality of illumination wavelengths, and a combined optical density of the reticulocyte corresponding to the plurality of illumination wavelengths.

42. The system of claim 40, wherein the electronic processor is configured to determine a mean reticulocyte hemoglobin value for the sample.