CN102599887B - Optical projection tomography method based on helical scanning track - Google Patents

Abstract

The invention discloses an optical projection tomography method based on a helical scanning track. The projection data obtained by an optical projection tomography (OPT) system in a helical scanning manner is rearranged to be transformed to a series of sinusoidal projection images scanned in a circular track, and the sinusoidal projection images are reconstructed, so as to obtain a sample three-dimensional fault structure. The optical projection tomography method provided in the embodiment of the invention can be used for effectively expanding the view of the optical projection tomography, in particular to the imaging view for long and thin objects.

Description

An Optical Projection Tomography Method Based on Helical Scanning Track

technical field

The present invention relates to an optical projection tomography (Optical Projection Tomography, referred to as OPT) technology, in particular to an optical projection tomography method based on a spiral scanning track.

Background technique

Optical projection tomography technology uses the characteristics of light propagating in a straight line in a small-sized organism, emits visible light to penetrate the sample, and then uses a camera to collect projected views of the sample from multiple angles for three-dimensional imaging. Specifically, when performing optical projection tomography, the sample needs to be scanned from multiple angles. Generally, an electronically controlled turntable is used to rotate the sample step by step, and one or more projection images are collected for each rotation to an angle. The sample does not move vertically and horizontally, but only rotates. From another angle, it can also be considered that the laser and the detector are rotating around the sample under test along a circular orbit. This scanning method is called circular orbit scanning. The final data collected by the circular orbit scanning of the optical projection tomography system is a series of two-dimensional projection images of light passing through the sample at different angles. If a certain line of all the projection images is extracted, they will be superimposed line by line according to the scanning sequence. A sinusoidal image similar to a sinusoidal curve can be obtained. Each sinusoidal image corresponds to a horizontal reconstruction slice of the sample, and all sinograms correspond to the three-dimensional tomographic reconstruction of the sample. From the projection data to the three-dimensional fault structure of the sample The process is called optical projection tomography for 3D reconstruction.

Optical projection tomography technology can realize structural and molecular-specific functional imaging of biological samples at a scale of 1-10 mm. It has many advantages such as high resolution, integrated structure and function, no radiation, and low cost. Carry out qualitative and quantitative research at the cellular level to achieve real-time, non-invasive, dynamic, and in vivo imaging of organisms. However, optical projection tomography is usually based on circular orbit scanning, and its imaging field of view is a cube with a limited field of view. Especially when scanning slender objects, it either uses a small optical path magnification to make the cube imaging field of view completely cover the entire sample. But the spatial resolution is poor; either a larger optical path magnification is used, and the spatial resolution is higher, but the imaging field of view cannot completely cover the sample, and only local fine imaging of the sample can be performed. At present, when scanning slender objects in the world, it is still impossible to achieve full sample coverage and high spatial resolution at the same time. This problem is called the "long object" problem.

The key reason for the problem that "long objects" cannot be imaged with high resolution in optical projection tomography is that the scanning process generally adopts a circular orbit, that is, the sample only rotates without translation. The effective imaging field of view of optical projection tomography is a cube due to circular orbit scanning , and for slender samples, which are not in the shape of a cube, in the optical projection tomography scan, if the cubic field of view is forced to completely cover the sample, the imaging accuracy will inevitably be lost in the length direction of the sample.

Contents of the invention

(1) Technical problems to be solved

In order to solve the problem that optical projection tomography circular orbit scanning cannot finely image long objects, the present invention provides an optical projection tomography method based on helical scanning orbit, which improves the optical projection tomography by performing three-dimensional imaging on the projection data of helical scanning. Imaging Field of view, accuracy and speed of imaging long objects.

(2) Technical solutions

The present invention performs three-dimensional tomographic imaging on the spiral trajectory projection data collected by the optical projection tomography system. By using the spiral scanning method, this embodiment expands the field of view of the optical projection tomographic imaging from a cube to a cuboid, effectively improving the axial imaging accuracy and solving the problem. Solved the problem that slender objects cannot be imaged with high resolution.

The present invention provides an optical projection tomography method based on a helical scanning track, which is characterized in that it includes:

Aiming at a series of projection images obtained by scanning the spiral orbit, the imaging field of view of the three-dimensional reconstruction is determined by using the axial position and projection angle of the projection images;

Dividing the three-dimensional reconstruction volume into a plurality of axial faults to be reconstructed, rearranging the projection rows corresponding to each axial fault to be reconstructed, and obtaining a sinogram corresponding to the axial faults to be reconstructed;

For each of the sinograms, fast tomographic reconstruction is performed using a hardware parallel method of a general-purpose graphics card;

All the fast tomographically reconstructed slices are stacked together sequentially to obtain a 3D reconstructed body. Specifically, the present invention includes two steps: data rearrangement and three-dimensional reconstruction, and these two steps can completely realize the three-dimensional tomographic imaging of the helical scanning data. Among them, the data rearrangement step rearranges a series of projection images obtained by spiral scanning according to the axial position and projection angle of each projection image, and extracts and stitches together the projection data corresponding to each fault to be reconstructed , to form a sinogram. Through this step, the spiral projection data can be converted into circular orbit projection data (in optical projection tomography, all the projection data of a radial fault to be reconstructed are combined into one image according to the angle, usually called is a sinogram, sinogram, which corresponds to the data required for reconstruction of the fault to be reconstructed); the three-dimensional reconstruction step uses the rearranged sinogram, uses the hardware parallel method of the general graphics card and the circular orbit filter back projection reconstruction method, and reconstructs The three-dimensional tomographic internal structure of the sample is constructed.

(3) Beneficial effects

The embodiment of the present invention can rapidly realize the three-dimensional reconstruction of the optical projection tomography helical scanning data, and expand the imaging field of view under the premise of ensuring the imaging accuracy.

Description of drawings

Fig. 1 is a schematic diagram of the scanning orbit of the imaging system in the optical projection tomography method based on the spiral scanning orbit in the embodiment of the present invention; wherein, Fig. 1a shows the circular orbit scanning mode and imaging field of view of ordinary optical projection tomography; Fig. 1b shows The helical orbit scanning mode and imaging field of view adopted in the embodiment of the present invention are shown;

Fig. 2 is the process of data rearrangement in the optical projection tomography method based on the helical scanning trajectory according to the embodiment of the present invention; Fig. 2a shows that in the helical 3D reconstruction volume, a slice to be reconstructed is mapped to the corresponding projection lines in the two projection images The process; Figure 2b shows the sinogram obtained by rearranging the data of all the projection lines corresponding to the faults to be reconstructed in Figure 2a according to the projection order;

FIG. 3 is a three-dimensional tomogram obtained by performing spiral scanning on a mouse bone and reconstructing it using the method of this embodiment in the optical projection tomography method based on a spiral scanning track in an embodiment of the present invention;

Fig. 4 is the result of visualization of the three-dimensional reconstruction of the mouse bone in the optical projection tomography method based on the helical scanning orbit in the embodiment of the present invention. It can be seen that the present invention can effectively expand the field of view of the optical projection tomography.

Detailed ways

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with specific embodiments and with reference to the accompanying drawings. While illustrations of parameters including particular values may be provided herein, it should be understood that parameters need not be exactly equal to the corresponding values, but rather may approximate the values within acceptable error margins or design constraints.

The invention is an optical projection tomography method based on a helical scanning track. The invention is specially aimed at the projection data collected by an optical projection tomography system in a spiral orbit scanning mode, and performs three-dimensional tomography imaging to expand the imaging field of view. As shown in Figure 1a, at present, optical projection tomography generally adopts a circular orbit scanning method, and its effective imaging field of view is a cube, but for slender samples, which are not in the shape of a cube, when performing optical projection tomography scanning, if forced When the sample is completely in the cube field of view, the sample will inevitably lose imaging accuracy in the axial direction; as shown in Figure 1b, the embodiment of the present invention adopts a new helical scanning method, using sample rotation and axial translation to image the field of view of optical projection tomography Expanding to a cuboid not only retains high precision, but also expands the imaging field of view. The invention is specially aimed at performing large-field imaging on scanning data of slender objects, and has the characteristics of fast, robust and high resolution.

The implementation scheme of the present invention is divided into two main steps: data rearrangement and three-dimensional reconstruction, wherein the data rearrangement uses the axial position and projection angle of the projection view to obtain a series of sinusoidal images, each sinusoidal image corresponds to a fault to be reconstructed, all The sinogram corresponds to the 3D reconstruction volume; fast 3D tomographic reconstruction uses a high-performance general-purpose graphics card to perform hardware parallel acceleration on the traditional circular orbit filter back-projection method, uses the sinogram to quickly reconstruct each fault, and stacks all the faults together one by one. Get a 3D reconstruction. The steps of the present invention are described below using the slender bone experiment. In the experiment, the bone of the mouse is used to carry out spiral optical projection tomography. The number of projection view pixels collected by the imaging system is 500*500, and the size of each pixel is 24 microns. The data acquisition process , the sample is rotated at an equal angle, and each time an angle is rotated, an axial equal-step translation is performed, so that the actual scanning track is a helical line, and the sample rotates for 3 times in total, that is, 3*360°=1080°, and the sample axis A total of 15 mm translation was performed, and a total of 1030 projected views were collected from angles. Taking the bone experiment as an example, the detailed steps of the present invention are as follows:

Step S1: In this step, for a series of axially shifted projection images collected in the spiral scanning mode, according to the axial position and scanning angle of each projection image, first use the axial positions of the first and last projection images , combined with the size of the projection image, calculate the imaging field of view of the entire reconstructed body; then perform the fault division of the reconstructed body, set the size of the fault pixel and the thickness of the fault as the pixel size of the projection graph, and calculate the number of voxels of the reconstructed body; then, For each fault to be reconstructed, its corresponding axial position is calculated, and its projection line in each projection map is located, and all corresponding projection lines are rearranged into a sinogram according to the scanning order. Figure 2 shows the process of spiral data rearrangement, in which Figure 2a shows the corresponding relationship between a fault to be reconstructed and the projection map in the 3D reconstruction volume, the figure shows two projection maps of 0° and 180°, and the The red line is the projection line corresponding to the fault to be reconstructed; Figure 2b shows the sinogram obtained after rearranging all the projection lines corresponding to the fault to be reconstructed in Figure 2a, and the corresponding fault in Figure 2a can be accurately reconstructed by using the sinogram.

本实施例中样品的螺旋扫描轨迹一般是沿着z轴向上，故

则投影图记录的最小轴向位置

投影图记录的最大轴向位置

样品轴向成像视野(field of view，FOV)为

即轴向总位移加上一个投影图的高度，径向成像视野一般与投影图的宽度相同，即Fov_x＝Fov_y＝N×d_s；重建体的体素大小与投影图的像素大小相同为d_s，故三维断层重建体的体素个数为Num_x＝Fov_x/d_s＝N，Num_y＝Fov_y/d_s＝N，Num_z＝Fov_z/d_s，其中Num_z表示三维重建体的断层数。本步骤可以确定待重建三维体的视野和体素数。Step S1-1: In order to carry out 3D reconstruction, it is necessary to first calculate the 3D imaging field of view, that is, the size of the 3D reconstruction body. Assume that the number of sample projection images collected is M, and the number of pixels in each projection image is N×N, where N is the projection image. The number of rows and columns, the pixel size of the projection image is d _s , the unit is mm, the length and width of the projection image are both N×d _s , the projection angle of the i-th projection image is r _i , where i=1, 2, 3 ,..., M, the axial position recorded during scanning of the i-th projection image is

The helical scanning trajectory of the sample in this embodiment is generally upward along the z axis, so

Then the minimum axial position recorded in the projection map is

The maximum axial position recorded by the projection diagram

The sample axial imaging field of view (field of view, FOV) is

That is, the total axial displacement plus the height of a projection image, the radial imaging field of view is generally the same as the width of the projection image, that is, Fov _x = Fov _y = N×d _s ; the voxel size of the reconstructed volume is the same as the pixel size of the projection image is d _s , so the number of voxels of the 3D tomographic reconstruction volume is Num _x =Fov _x /d _s =N, Num _y =Fov _y /d _s =N, Num _z =Fov _z /d _s , where Num _z represents The number of slices in the 3D reconstruction. In this step, the field of view and the number of voxels of the three-dimensional body to be reconstructed can be determined.

为第1个断层底平面对应的z轴位置，对于第k个断层，该层的中心平面对应第(k-0.5)个断层，0.5代表了半个断层，(k-0.5)×d_s代表了第k个断层中心平面相对于第1个断层底平面在z轴上的距离。对于第i个投影图，其所覆盖的z轴区间为

其中

两者相差一个投影图的高度。若

说明第k个断层可以投影到第i个投影图上，本发明实例则找出第k个待重建断层对应第i个投影图中最近的一个行

其中[]为取整符号，

即第k重建断层在第i个投影图中的投影行，如图2a所示，投影图上的红线表示断层对应的投影行；我们将第k个待重建断层对应的所有投影行按照扫描顺序排列为一幅正弦图，即完成第k个断层的投影数据重排，图2b显示了图2a中待重建断层的正弦图，本步骤将依次完成所有断层的数据重排。Step S1-2: The spiral 3D reconstruction process can be transformed into a series of fault reconstructions in the z-axis direction. In this step, through data rearrangement, the sinogram corresponding to each fault to be reconstructed in the z-axis direction is found, and then the faults are reconstructed one by one. This embodiment first calculates the z-axis position corresponding to each fault to be reconstructed, uses this position to find its corresponding projection line in each projection map, and then arranges all projection lines of the fault into a sinogram according to the scanning order , that is to complete the rearrangement of the data corresponding to the fault. The details are as follows: For the kth fault to be reconstructed (where k=1, 2, 3, ..., Num _z ), the position of its central plane in the z-axis direction In this embodiment, the z-axis position corresponding to the fault center plane is used as the z-axis position of the fault,

is the z-axis position corresponding to the bottom plane of the first fault. For the kth fault, the central plane of this layer corresponds to the (k-0.5)th fault, 0.5 represents half of the fault, and (k-0.5)×d _s represents is the distance on the z-axis between the center plane of the kth fault and the bottom plane of the first fault. For the i-th projection map, the z-axis interval covered by it is

in

The difference between the two is the height of a projection map. like

It shows that the kth fault can be projected onto the i-th projection map, and the example of the present invention finds out that the k-th fault to be reconstructed corresponds to the nearest row in the i-th projection map

Where [] is rounding symbol,

That is, the projection line of the k-th reconstruction fault in the i-th projection map, as shown in Figure 2a, the red line on the projection map indicates the projection line corresponding to the fault; we put all the projection lines corresponding to the k-th fault to be reconstructed in the scanning order Arranging into a sinogram means completing the rearrangement of the projection data of the kth fault. Figure 2b shows the sinogram of the fault to be reconstructed in Figure 2a. This step will complete the data rearrangement of all faults in sequence.

Step S2: performing tomographic reconstruction on each rearranged sinogram one by one to obtain a three-dimensional tomographic reconstruction. This step adopts the parallel circular orbit filter back projection reconstruction method, and those skilled in the art should understand that the circular orbit filter back projection reconstruction method is one of the most classic reconstruction methods in the field of circular orbit optical projection tomography, which includes filtering and back projection Two steps, this step will be filtered on the CPU, and the back projection has a relatively large amount of calculation and has the characteristics of a high degree of parallelism, and the present invention performs parallel acceleration on a high-performance general-purpose graphics card to realize parallel circular orbit filtering The back projection reconstruction method improves the reconstruction speed. Compared with the CPU, the high-performance graphics card has more stream processors, and can accelerate a large number of similar independent calculations in parallel. However, in the embodiment of the present invention, in the process of back-projection from the filtered sinogram to the slice, each pixel of the slice is The reconstructions (i.e. backprojections) are independent and fully parallel. This embodiment implements the hardware parallel back-projection method in the graphics card, first the sinogram after filtering is copied from the memory of the computer to the video memory of the graphics card, and then a plurality of stream processors of the graphics card are used to process the sinogram in the video memory Perform parallel back-projection operation, each stream processor independently performs a back-projection operation of a fault pixel, each time a pixel back-projection is completed, the stream processor turns to other unprocessed pixels to continue back-projection, until all pixels of the fault are in the video memory After the back-projection operation is completed, the reconstruction result of the fault can be obtained. In this embodiment, the fault result in the display memory is copied to the internal memory for display and storage.

For the helical scanning three-dimensional reconstruction method of the present invention, Fig. 3 is a three-dimensional tomogram obtained by using the mouse bone for spiral scanning reconstruction in the optical projection tomography method based on the helical scanning orbit in the embodiment of the present invention, and the internal structure of the bone can be seen very clearly.

Fig. 4 is the result of visualizing the three-dimensional structure of the bone in the optical projection tomography method based on the helical scanning track in the embodiment of the present invention. It can be seen that the imaging field of view completely covers the elongated bone through the helical reconstruction, which further verifies the present invention. validity of the invention.

To sum up, the present invention proposes an optical projection tomography method based on a helical scanning orbit to realize three-dimensional tomographic imaging of helical scanning projection data, which has the characteristics of large imaging field of view, high precision, and fast imaging. Specifically:

(1) In terms of scanning mode, the embodiment of the present invention realizes the helical scanning mode of optical projection tomography and expands the imaging field of view based on the traditional circular orbit parallel beam scanning and the axial translation of the sample;

(2) In the process of data rearrangement, the embodiment of the present invention aims at the helical scan projection map, uses the scan angle and axial position of the projection map to find out the sinogram corresponding to each fault to be reconstructed, and rearranges the helical scan data The form of circular orbit scanning data not only simplifies the reconstruction process, but also ensures the imaging accuracy;

(3) In the process of spiral three-dimensional reconstruction, the embodiment of the present invention performs parallel acceleration on the traditional circular orbit filter back projection algorithm, which improves the speed of three-dimensional tomographic imaging.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (3)

1. An optical projection tomography method based on a spiral scanning track is characterized by comprising the following steps:

aiming at a series of projection drawings obtained by helical track scanning, calculating the imaging visual field of the three-dimensional reconstruction body according to the following formula by using the axial position and the projection angle of the projection drawings:

Fou_x=Fou_y=N×d_s

$\mathrm{Fo}{u}_z=z_{\max }^p-z_{\min }^p+N\×d_s$

wherein Fou_xAnd Fou_yFor radial imaging views, i.e. imaging views in the horizontal direction, in mm, Fou_xIndicating the position in the horizontal direction on the x-axis, Fou_yIndicating the position in the horizontal y-axis, Fou_zIn the axial imaging field, i.e. the imaging field of view of the vertical z-axis, which is measured in millimeters, N is the number of rows and columns of the projection map, in the helical scanning, each time one projection map is scanned, the imaging system records the corresponding z-axis position of the projection map,

is the maximum value of the z-axis position of the projection image in the spiral scanning,

is the minimum value of the z-axis position of the projection image in millimeters during helical scanning_sThe maximum value and the minimum value of the z-axis position of the projection graph are calculated according to the following formula:

$z_{\min }^p=z_1^p$

$z_{\max }^p=z_M^p$

where M is the number of projected views of the sample,

i =1,2, M, the scanning spiral trajectory is along the z-axis for the axial position recorded at the scanning of the ith projection image, so

And

first and last projection respectively;

dividing the three-dimensional reconstruction body into a plurality of axial to-be-reconstructed faults, and rearranging the data of the projection line corresponding to each axial to-be-reconstructed fault to obtain a sinogram corresponding to the axial to-be-reconstructed fault: wherein the number of reconstructed volume pixels is calculated according to the following formula:

Num_x=Fou_x/d_s=N

Num_y=Fou_y/d_s=N

Num_z=Fou_z/d_s

wherein Num_xNumber of lines, Num, for each fault to be reconstructed_yNumber of lines, Num, for each fault to be reconstructed_zThe voxel size of the reconstructed volume is d for the number of faults in the z-axis direction_sThe same size as the pixels of the projected image;

when data rearrangement is carried out, a projection line corresponding to each fault to be reconstructed in a projection diagram is calculated, and if the kth fault to be reconstructed is projected to the ith projection diagram, the projection line corresponding to the fault is calculated according to the following formula:

${\mathrm{Raw}}_i^k=[(z_k^{\mathrm{slice}}-z_i^{\mathrm{pFouMin}})/d_s)]+1$

wherein k =1,2, 3, ·, Num_z，

For the z-axis position corresponding to the kth fault,

for the minimum z-axis position covered by the ith projection,

recording a z-axis position during scanning of the ith projection drawing, wherein the recorded position during scanning is the z-axis position corresponding to the bottom of the projection drawing; wherein the z-axis position corresponding to the kth reconstructed fault is calculated according to the following formula:

$z_k^{\mathrm{slice}}=(k-0.5)\×d_s+z_{\min }^p$

wherein k =1,2, 3.. times.num_zTaking the z-axis position corresponding to the central plane of the fault as the z-axis position of the fault,

for the z-axis position corresponding to the bottom plane of the 1 st fault, for the k-th fault, the central plane of the slice corresponds to the (k-0.5) th fault, 0.5 represents a half fault, (k-0.5) x d_sRepresents the distance of the k-th fault center plane relative to the 1 st fault bottom plane on the z-axis; if the kth fault to be reconstructed is projected to the ith projection diagram, the kth fault to be reconstructed satisfies the following formula:

$z_i^{\mathrm{pFouMin}}<z_k^{\mathrm{slise}}<z_i^{\mathrm{pFouMax}}$

wherein,

the minimum z-axis position covered by the ith projection image is the position recorded when the projection image is scanned;the maximum z-axis position covered by the ith projection image, which is the minimum position plus the height of one projection image;

aiming at each sinogram, performing rapid fault reconstruction by using a hardware parallel method of a general graphic card, which specifically comprises the following steps: firstly copying the filtered sinogram from a memory of a computer to a display memory of a graphic card, then carrying out parallel back projection operation on the sinogram in the display memory by using a plurality of stream processors of the graphic card, wherein each stream processor independently carries out back projection operation of a fault pixel, and when the back projection of one pixel is finished, the stream processors turn to other unprocessed pixels to continue back projection until all pixels of the fault are completely back projected in the display memory, so that the reconstruction result of the fault can be obtained;

and overlapping all the faults reconstructed by the rapid fault in sequence to obtain the three-dimensional reconstructed body.

2. The helical scan trajectory-based optical projection tomography method as claimed in claim 1, wherein during the data rearrangement, for each slice to be reconstructed, all projection lines corresponding thereto are found, and the projection lines are superimposed into a sinogram according to the scan order, and then the data rearrangement of the sinograms corresponding to the slices is completed one by one along the z-axis direction.

3. The helical scanning trajectory-based optical projection tomography method as claimed in claim 1, wherein in the three-dimensional reconstruction, for each sinogram, each tomographic plane is reconstructed by using a parallel circular trajectory filtering back-projection reconstruction method, and a three-dimensional tomographic reconstruction volume is obtained.

Images