CN107705336B - A method for adjusting the coloring components of pathological images - Google Patents

Abstract

The invention discloses a method for adjusting the dyeing components of pathological images. First, the brightness and saturation of the pathological slices are corrected in the hue, saturation, and brightness (HSV) space of the collected pathological slices, and the saturation of the pathological slices is corrected by correcting the saturation of the pathological slices. Degree and brightness, separation of dyeing components, adjustment of dyeing components, and synthesis of dyeing components, realize the algorithm of adjusting the content of a single dyeing component, achieve the effect of independent adjustment of different dyes in pathological images, and effectively complete the task of dyeing separation , to assist in the diagnosis of pathologists.

Description

A method for adjusting the coloring components of pathological images

technical field

The invention relates to the field of digital image processing, in particular, to a method for adjusting dyeing components of pathological images.

Background technique

Digital pathological images are high-resolution digital images obtained by scanning and collecting pathological sections through an automatic microscope or optical magnification system, which have been widely used in pathological clinical diagnosis. The color of pathological sections is obtained by staining with dyes, and the most commonly used staining method is hematoxylin and eosin (H-E) staining. However, human factors such as manual operation techniques and differences in the ratio of dyes during the staining process will cause differences in the staining quality of pathological sections; at the same time, the difference in lighting environment during the section scanning process also makes the collected digital pathological images Brightness There is a big difference with saturation. These differences will hinder the pathologist's diagnosis and affect the judgment of diagnostic accuracy, so it is necessary to perform color correction on digital pathological images.

With the continuous development of digital image processing technology, some methods for color enhancement and correction applied to natural scene images have matured, and are widely used in vision and multimedia systems, biomedicine, industrial engineering, aerospace and other fields, including histogram equalization Algorithm, Retinex algorithm and enhancement method based on color space transformation, etc. The general flow of these algorithms is shown in Figure 1.

The histogram equalization algorithm is an adaptive enhancement method. The output result is the optimal effect calculated by the algorithm. The doctor cannot adjust the result. The adjustment method corresponding to the histogram equalization method is the histogram regulation. In contrast, Retinex algorithm and enhancement algorithm based on spatial transformation are enhancement methods with adjustable parameters. All three methods can artificially adjust the color of the pathological image. However, the color of the pathological image is obtained by coloring with two or more dyes, and the color corresponding to the dye has a specific meaning, such as hematoxylin. The stain can stain chromatin and ribosomes in the cytoplasm to violet blue, and the eosin stain can stain the components in the cytoplasm and extracellular matrix red. What the doctor needs in the diagnosis process is to adjust the components of each dye individually to enhance or weaken the intensity of the dye without affecting other dyes. The above methods are all designed for natural images stored in red, green and blue (RGB) channels. When applied to pathological images, adjusting a single channel will inevitably affect each coloring component, which cannot be very good according to the needs of doctors. Complete the task of adjusting the staining of pathological images.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for adjusting the coloring components of a pathological image, so as to solve the problem that the existing image processing technology cannot separate the different coloring agents in the pathological image and realize the independent adjustment of the different coloring agents in the pathological image.

For realizing the purpose of the present invention, adopt following technical scheme:

A method for adjusting the dyeing components of a pathological image, comprising the following steps:

(1). Collect pathological slices: collect the pathological slices into the computer, which are represented by RGB channels, where the pixel coordinates are marked as (x, y);

(2). Correcting the saturation and brightness of the pathological slice, including the following three steps:

a. Transform the pathological section from the RGB channel to the HSV channel;

b. Delineate the pixel with the lowest 5% of the saturation S channel value of the pathological slice as the background area pixel; count the mean value of the background area pixels to estimate the saturation of the background area, and express it as S _back , and at the same time count all the pixels in the background area. The mean value of the pixels of the background area in the luminance channel V, as the luminance value of the background area, and expressed as V _back ; Afterwards, with the background area being transformed into white as the target, the saturation of the entire described pathological slice is linearly stretched and brightness, while keeping the hue unchanged;

c. Inversely transform the enhanced pathological image in step b to the RGB channel to complete the correction of the saturation and brightness of the pathological image;

的染色分离；涉及公式为：(3). Separation of dyeing components: in the RGB channel transformed in step 2, the coloring by the optical density O _c (x, y) of the channel c (c=R, G, B) and the coloring agent s The intensity A _s (x, y) obtains the mapping relationship between the optical density and the coloring intensity of the dye s, and using the mapping relationship, the color deconvolution algorithm is used to complete the image

The dye separation of ; the formula involved is:

Wherein, the A ₀ is the maximum value of the coloring strength of the dye, and the A ₀ =1;

(4) Adjustment of dyeing components: after obtaining the coloring intensity A _s '(x, y) of each dye, adjust it according to the needs of diagnosis; let the adjustment rate of the dye s be p _s , where all Said p _s > 0, the formula for calculating the tinting strength of the adjusted dye is:

Wherein, the p _s >1 represents enhancing the tinting strength of the dyeing component s, and p _s <1 represents reducing the tinting intensity of the dyeing component s;

(5) Synthesis of dyeing components: After adjusting the dyeing components in the step (4), the dyeing data is fused and inversely transformed back to RGB channels.

In the above-mentioned method for adjusting the staining components of pathological images, preferably, the numerical values of the R, G, and B channels of the pixel coordinates (x, y) in step 1 are expressed as:

I(x,y)=[ _Ir (x,y),Ig(x,y), _Ib ( _x ,y)] (3)

where I _r (x, y), I _g (x, y), and I _b (x, y) represent the values of the three color channels of red, green and blue, respectively, and I _c (x, y)∈[0,1] ,c=r,g,b.

In the above-mentioned method for adjusting the dyeing components of pathological images, preferably, when the channel is changed in the step a, the formula involved in the transformation is as follows:

Wherein, the H(x,y), S(x,y) and V(x,y) respectively represent the hue, saturation and brightness of the pixel point (x,y).

且背景亮度

涉及计算公式如下：In the above-mentioned method for adjusting the staining components of pathological images, preferably, in the step b, the saturation and brightness of the linearly stretched entire pathological slice refer to the background saturation of the processed pathological slice

and background brightness

The calculation formula involved is as follows:

和

代表点(x,y)的增强结果；Among them, the

and

The enhancement result of the representative point (x, y);

In the step c, the transformation formula involved is as follows:

表示

的整数部分，所述

表示饱和度、亮度校正后的病理图像中点(x,y)的值。如上所述的病理图像染色成分调节方法，优选地，通道c(c＝r,g,b)的所述光密度O_c(x,y)的计算公式如下所示：Among them, the

express

the integer part of the

Indicates the value of the point (x, y) in the pathological image after saturation and brightness correction. In the above-mentioned method for adjusting the staining components of pathological images, preferably, the calculation formula of the optical density O _c (x, y) of the channel c (c=r, g, b) is as follows:

Wherein, the I _0,c is the maximum value of a single channel, and the I _0,c =1.

In the above-mentioned method for adjusting the coloring components of pathological images, preferably, in the step (3), when the coloring agent is colored by a single coloring agent, the optical density O _c (x, y) is related to the coloring degree A of the coloring agent. In direct proportion, when the dye is colored by multiple dyes, the optical density is equal to the sum of the optical densities of each of the dyes on channel c. When the dye is hematoxylin-eosin-diaminobenzidine, the Described hematoxylin is denoted as H, described eosin is denoted as E, described diaminobenzidine is denoted as DAB, its described optical density O _c (x, y) and the tinting strength of dyeing agent As ( _x , y ) and the conversion relationship is as follows:

代表所述染色剂s在通道c上的吸光率，

对于染色剂s和通道c是一个常量，可以经由单一所述染色剂染色测试得到，所述染色剂为H-E-DAB染色时，通道c对三种所述染色剂H,E和DAB的吸光率矩阵为Wherein, the s=H, E, DAB,

represents the absorbance of the dye s on channel c,

For the dye s and channel c is a constant, which can be obtained through the dyeing test of a single said dye, when the dye is HE-DAB staining, the absorbance of channel c to the three dyes H, E and DAB The matrix is

make

O=[O _r (x,y),O _g (x,y),O _b (x,y)] ^T ,

A=[A _H (x,y),A _E (x,y),A _DAB (x,y)] ^T ,

Formula (8) is abbreviated as:

O=M·A (9)

Let D=M ^-1 , from the formula (7), the tinting strength of each of the dyes can be obtained as:

A = D · O

(10)

Among them, D is called the color deconvolution matrix, which represents the mapping relationship between the optical density and the coloring intensity of the dye. When the dye is H-E-DAB dyeing, the deconvolution matrix is:

的染色分离。The vector A=[A _H (x, y), A _E (x, y), A _DAB (x, y)] ^T is the decomposed staining intensity. Since the A is a linear transformation of the optical density O, so The A is still in optical density space, inversely transform it into linear space to complete the image

staining separation.

In the above-mentioned method for adjusting the staining components of pathological images, preferably, in the step 5, when the staining is hematoxylin-eosin-diaminobenzidine staining, the calculation formula used for fusion of the staining data is as follows :

为调节之后的结果。Calculated

is the result after adjustment.

The invention provides a method for adjusting the dyeing components of a digital pathological image. First, the brightness and saturation of the pathological image are corrected in the hue, saturation, and brightness (HSV) space, and then the color deconvolution algorithm is used to realize the adjustment of a single dyeing component. The algorithm that adjusts the content of , achieves the effect of independent adjustment of different dyes in pathological images, effectively completes the task of dye separation, and assists pathologists in diagnosis.

Description of drawings

FIG. 1 is a flow chart of digital image enhancement in the prior art.

Figure 2 is a flow chart of the method of the present invention.

FIG. 3 is an effect diagram of four pathological sections after processing in the preferred embodiment 1 of the present invention.

FIG. 4 is the original image in the preferred embodiment 2 of the present invention.

FIG. 5 is an effect diagram after adjustment of the original FIG. 4 in the preferred embodiment 2 of the present invention.

Detailed ways

In the prior art, the general pathological image adjustment method is carried out in the RGB channel or the HSV channel, and the staining information of each dye will be distributed in the RGB channel or the HSV channel at the same time, while the general pathological image adjustment method can only be used in the RGB space. Or adjusting a single channel in HSV space, two problems are involved here: 1) Adjusting a channel in RGB space or HSV space alone will affect each coloring component at the same time; 2) Want to adjust one coloring component Three channels of RGB space or HSV space need to be scaled simultaneously. Therefore, it is difficult for general pathological image adjustment methods to adjust a single stained component.

In the present invention, the color deconvolution algorithm is used to transform the pathological image into the dye space, that is, after the transformation, each dye is controlled by a single channel, so that a single dye component can be realized by adjusting the value of a certain channel after the transformation. adjustment without affecting other dyeing components. The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

Example 1

A method for adjusting the dyeing components of a pathological image, the specific process is shown in Figure 2, and includes the following steps:

1. Collection of pathological sections

The pathological image is input into the computer through the pathological slice scanning equipment, and the image is represented in the RGB color space.

I(x,y)=[ _Ir (x,y),Ig(x,y), _Ib ( _x ,y)] (3)

where I _r (x, y), I _g (x, y), and I _b (x, y) represent the values of the three color channels of red, green and blue, respectively, and I _c (x, y)∈[0,1] ,c=r,g,b.

2. Saturation and Brightness Correction

When the pathological image is scanned, the saturation and brightness of the image will be poor due to poor lighting conditions and other reasons. Therefore, it is necessary to correct the brightness and saturation of pathological images. The background area (area without tissue coverage) of the pathological section does not contain any content. Generally, the imaging effect of the entire section is best when the background area is pure white. Based on this, this method realizes the correction of the saturation and brightness of the entire slice. The specific method includes the following three steps:

a. Transform the image I(x,y) from RGB space to HSV space, and the transformation formula involved is as follows:

Among them, H(x,y), S(x,y) and V(x,y) represent the hue, saturation and brightness of the point (x,y), respectively.

且背景亮度

同时保持色调不变，涉及计算公式如下：b. In the whole slice, the saturation of the background area is always the lowest. Therefore, the pixels with the lowest 5% value of the saturation S channel are defined as the pixels of the background area, and then the average value of these pixels is calculated to estimate the saturation of the background area, which is expressed as S _back , and these pixels are counted in the luminance channel V at the same time. The mean value, as the luminance value of the background area, and denoted as V _back . Finally, aiming to transform the background area into white, linearly stretch the saturation and brightness of the entire slice, that is, try to ensure the background saturation of the slice after processing.

and background brightness

While keeping the hue unchanged, the calculation formula involved is as follows:

和

代表点(x,y)的增强结果。in

and

Represents the augmented result for the point (x,y).

c. Inversely transform the enhanced result to RGB space to complete the correction of the saturation and brightness of the pathological image. The transformation formula involved is as follows:

表示

的整数部分，

表示饱和度、亮度校正后的病理图像中点(x,y)的值。in

express

the integer part of ,

Indicates the value of the point (x, y) in the pathological image after saturation and brightness correction.

The method described in step 2 was used to process four pathological slices, and the results are shown in Figure 3, where (a), (b), (c), and (d) are four digital pathological slices with poor imaging conditions, The left half of each slice is the original image, and the right half is the enhanced effect processed by the method of the present invention. It can be seen from the figure that the method of the present invention can effectively adjust the saturation and brightness of the pathological image.

3. Separation of dyeing components

In the RGB channel, the optical density calculation formula of channel c (c=r, g, b) is:

所以I_0,c＝1。在单独染色剂着色时，光密度O_c(x,y)与染色剂着色度A成正比，用多种染色剂着色时，光密度等于各染色剂在通道c上光密度的和，以H-E-DAB染色为例，光密度O_c(x,y)(c＝r,g,b)和染色剂的着色强度A_s(x,y)(s＝H,E,DAB)的转换关系如下：Among them, I _{0, c} is the single-channel maximum value, (in this method

So I _0,c =1. When coloring with a single colorant, the optical density O _c (x, y) is proportional to the coloring degree A of the colorant. When coloring with multiple colorants, the optical density is equal to the sum of the optical densities of each colorant on channel c, with HE - DAB dyeing as an example, the conversion relationship between the optical density O _c (x, y) (c=r, g, b) and the coloring intensity of the dye As (x, y) ( _s =H, E, DAB) is as follows :

代表染色剂s在通道c上的吸光率，

对于染色剂s和通道c是一个常量，可以经由单一染色剂染色测试得到，以H-E-DAB染色为例，通道c对三种染色剂H,E和DAB的吸光率矩阵为in,

represents the absorbance of dye s on channel c,

For the dye s and channel c, it is a constant, which can be obtained by a single dye staining test. Taking HE-DAB staining as an example, the absorbance matrix of channel c for three dyes H, E and DAB is:

Let O=[O _r (x,y),O _g (x,y),O _b (x,y)] ^T ,A=[A _H (x,y),A _E (x,y),A _DAB (x,y)] ^T , formula (6) can be abbreviated as:

O=M·A (9)

Let D=M ^-1 , from the formula (7), the tinting strength of each colorant can be obtained as:

A=D·O (10)

Among them, D is called the color deconvolution matrix, which represents the mapping relationship between the optical density and the coloring intensity of the dye. Taking H-E-DAB dyeing as an example, the deconvolution matrix is:

的染色分离，涉及公式为：The vector A=[A _H (x, y), A _E (x, y), A _DAB (x, y)] ^T is the decomposed staining intensity. Since A is a linear transformation of the optical density O, A is still In the optical density space, inversely transform it into a linear space to complete the image

The dye separation involving the formula is:

A ₀ is the maximum value of the coloring intensity of the dye, which is the value range ([0, 1]) of the corresponding RGB channel in this method, and A ₀ =1 is taken.

4. Adjustment of dyeing components

After obtaining the tinting intensity As'( _x ,y) of each stain, the doctor can adjust it according to the diagnostic needs. Let the adjustment rate of the dyeing agent be p _s , where the p _s > 0, and the formula for calculating the coloring intensity of the adjusted dyeing agent is:

where _ps > 1 represents the enhancement of the tinting strength of the dyeing component _s , and ps <1 represents the weakening of the tinting strength of the dyeing component s.

5. Synthesis of dyeing components

After adjusting the coloring components, the coloring data needs to be fused and inversely transformed back to RGB channels. Taking H-E-DAB staining as an example, the calculation formula is as follows:

即为调节之后的结果。Calculated

It is the result after adjustment.

. FIG. 5 shows the effect of processing the HE-stained digital pathological image by the staining adjustment method of the present invention. HE-staining is the most commonly used staining method, H stands for hematoxylin stain, which can stain cell nuclei blue-violet, and E represents eosin stain, which can stain matrix pink. HE-stained pathological images can be regarded as HE-DAB-stained images without DAB staining, so the deconvolution matrix provided by formula (11) can also be used for processing, just make p _DAB =1 in formula (13), Only adjust pH and _{pE .} The dyeing adjustment method at different _pH and _pE values is shown in Figure 5. It can be seen from the results that, using the method of the present invention, the two staining components of the pathological image can achieve the effect of independent adjustment.

The method of the present invention can be divided into two parts, the first part is the adjustment of pathological image saturation and brightness based on HSV space, the content of the second step; the second part is the color component adjustment based on color deconvolution, the third, the fourth, 5 steps of content. Among them, the first part is an adaptive image adjustment method, which does not require human intervention, and performs color correction on the entire pathological image, which is more targeted than other color correction methods. The second part is an image enhancement method for the doctor to manually set the parameter _ps according to the diagnosis needs, in order to adjust the coloring intensity of the dye. The previous image adjustment methods are all performed in RGB space, and it is difficult for users to obtain the effect of adjusting the dyeing density of a single dye by adjusting the brightness and saturation in RGB. The invention uses the color deconvolution algorithm to transform the pathological image into the dyeing space, then adjusts the coloring intensity of each dye in the dyeing space, and finally inversely transforms it into the RGB space to achieve the effect of adjusting a single dye.

Claims (6)

1. a pathological image staining component adjustment method, is characterized in that, comprises the following steps: (1). Collect pathological slices: collect pathological slices in computer, represent with RGB channel, and wherein pixel point coordinates are marked as (x, y) ); (2). Correcting the saturation and brightness of the pathological slice, including the following three steps: a. Transforming the pathological slice from the RGB channel to the HSV channel; b. Delineate the pixel with the lowest 5% of the saturation S channel value of the pathological slice as the background area pixel; count the mean value of the background area pixels to estimate the saturation of the background area, and express it as S _back , and at the same time count all the pixels in the background area. The mean value of the pixels of the background area in the luminance channel V, as the luminance value of the background area, and expressed as V _back ; Afterwards, with the background area being transformed into white as the target, the saturation of the entire described pathological slice is linearly stretched and brightness, while keeping the hue unchanged; c. Inversely transform the enhanced pathological image in step b to the RGB channel to complete the correction of the saturation and brightness of the pathological image; (3). Separation of dyeing components: in the RGB channel transformed in the step (2), the coloring by the optical density O _c (x, y) of channel c (c=R, G, B) and the coloring agent s The intensity A _s (x, y) obtains the mapping relationship between the optical density and the coloring intensity of the coloring agent s, and using the mapping relationship, the color deconvolution algorithm is used to complete the color separation of the image; the related formula is:

Wherein, the A ₀ is the maximum value of the coloring strength of the dye, and the A ₀ =1; (4). Adjustment of dyeing components: after obtaining the As '( _x , y) of the coloring intensity of each dye, adjust it according to the needs of diagnosis; let the adjustment rate of the dye s be p _s , where Said p _s > 0, the formula for calculating the tinting strength of the adjusted dye is:

Wherein, the p _s >1 represents enhancing the tinting intensity of the dyeing component s, and p _s <1 represents reducing the tinting intensity of the dyeing component s; (5) Synthesis of dyeing components: After adjusting the dyeing components in the step (4), the dyeing data is fused and inversely transformed back to RGB channels. 2. The method for adjusting the dyeing components of a pathological image as claimed in claim 1, wherein the numerical values of the R, G, B three channels of the pixel coordinates (x, y) described in the step (1) are expressed as: I(x,y)=[ _Ir (x,y),Ig(x,y), _Ib ( _x ,y)] (3) where I _r (x, y), I _g (x, y), and I _b (x, y) represent the values of the three color channels of red, green and blue, respectively, and I _c (x, y) ∈ [0, 1] , c=r, g, b. 3. The method for adjusting the coloring components of pathological images as claimed in claim 1, wherein when the channel is changed in the step a, the formula involved in the transformation is as follows: _Vmax =max[ _Ir (x,y),Ig(x,y), _Ib ( _x ,y)] _Vmin =min[ _Ir (x,y),Ig(x,y), _Ib ( _x ,y)]

Wherein, the H(x, y), S(x, y) and V(x, y) respectively represent the hue, saturation and brightness of the pixel point (x, y); In the step b, the saturation and brightness of the linearly stretched entire pathological slice refer to the background saturation of the processed pathological slice, and the background brightness involves the following calculation formula: The involved calculation formula is as follows:

Among them, the

and

The enhancement result of the representative point (x, y); In the step c, the transformation formula involved is as follows:

in,

express

The integer part of the

Indicates the value of the point (x, y) in the pathological image after saturation and brightness correction. 4. The method for adjusting the dyeing components of a pathological image according to claim 1, wherein in the step (3), the optical density O _c (x, The calculation formula of y) is as follows:

Wherein, the I _0,c is the maximum value of a single channel, and the I _0,c =1. 5. The method for adjusting the coloring components of pathological images according to claim 1, wherein when the coloring agent is a single coloring agent in the step (3), the optical density O _c (x, y) It is proportional to the coloring degree A of the dye. When the dye is colored by multiple dyes, the optical density is equal to the sum of the optical densities of each of the dyes on channel c; when the dye is hematoxylin In the case of sperm-eosin-diaminobenzidine, hematoxylin is recorded as H, eosin is recorded as E, and diaminobenzidine is recorded as DAB, and its optical density is O _c (x, y) (c=r, g , b) and the tinting strength As ( _x , y) of the dye are converted as follows:

where the s=H, E, DAB, m _c ^s represents the absorbance of the dye s on channel c, m _c ^s is a constant for the dye s and the channel c, which is passed through a single said dye The dyeing test shows that when the dye is HE-DAB dyeing, the absorbance matrix of channel c to the three dyes H, E and DAB is:

make O=[O _r (x,y),O _g (x,y),O _b (x,y)] ^T ,A=[A _H (x,y),A _E (x,y),A _DAB (x,y)] ^T , formula (8) is abbreviated as: O=M·A (9) Let D=M ^-1 , from the formula (7), the tinting strength of each of the dyes can be obtained as: A=D·O (10) Among them, D is called the color deconvolution matrix, which represents the mapping relationship between the optical density and the coloring intensity of the dye. When the dye is H-E-DAB dyeing, the deconvolution matrix is:

The vector A=[A _H (x, y), A _E (x, y), A _DAB (x, y)] ^T is the decomposed staining intensity. Since the A is a linear transformation of the optical density O, so The A is still in the optical density space, and it is inversely transformed into a linear space to complete the color separation of the image. 6 . The method for adjusting the staining components of a pathological image according to claim 1 , wherein in the step (5), when the staining agent is hematoxylin-eosin-diaminobenzidine staining, the The calculation formula used for the fusion of staining data is as follows:

Calculated

is the result after adjustment; wherein, I _0,c is the maximum value of a single channel, and the I _0,c =1.

Images