Automatic Segmentation Method, Device, Computing Device and Storage Medium for Coronary OCT Images

Abstract

The present invention discloses an automatic segmentation method, device, computing device and storage medium for coronary OCT images, including: adopting a rough segmentation and a fine segmentation process for the OCT images, and using the fine segmentation result of the previous frame of OCT image as the rough segmentation result of the current frame of OCT image to continue the fine segmentation process, ensuring the continuity of the overall segmentation result of the lumen boundary in the OCT image sequence; in this way, for the OCT images with blood residue in the coronary lumen, based on the fine segmentation result of the previous OCT image without blood residue for further fine segmentation, the influence of the residual blood on the lumen segmentation accuracy can be greatly reduced, and the segmentation accuracy of the OCT images of blood can be significantly improved; using the dijkstra minimum path algorithm to search and identify the lumen boundary in the image can accurately distinguish the branch vessels and avoid inaccurate lumen boundary segmentation caused by overestimating the lumen at the bifurcation opening of the blood vessels.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing, and particularly relates to an automatic segmentation method, device, computing device and storage medium for coronary OCT images. Background Art

[0002] In current clinical practice, the diagnosis of coronary artery disease mostly uses invasive imaging methods, among which coronary angiography is the most popular one and is currently considered the gold standard for cardiac catheterization and hemodynamic assessment. However, the luminogram generated by coronary angiography only depicts the two-dimensional projected diameter of the lumen at a specific projection angle, without any geometric morphology information about the lumen or plaque. Therefore, recent coronary angiography is often accompanied by newer intravascular imaging techniques such as Intra Vascular UltraSound (IVUS) and Optical Coherence Tomography (OCT), which use ultrasonic waves and near-infrared light respectively to generate cross-sectional images of the coronary artery. OCT images have extremely high image contrast and spatial resolution (10 - 20 microns), and their spatial resolution can often reach about 10 times that of IVUS images. Therefore, they can clearly characterize the morphology of atherosclerotic plaques and the stent implantation status, including stent apposition and stent coverage, etc.

[0003] Parameters such as the minimum lumen diameter, minimum lumen area, and stenosis degree obtained by analyzing OCT images are of great significance for the diagnosis of coronary artery disease. Therefore, in OCT image analysis, the accurate segmentation of the coronary artery lumen is one of the main challenges.

[0004] When performing OCT imaging, it is necessary to empty the blood in the coronary artery within a certain period of time so as to obtain more clear structural information of the blood vessel wall, thereby performing lumen segmentation or plaque recognition. However, in practical applications, there is often a certain amount of blood residue in the imaged vascular segments of coronary OCT (especially at both ends), which causes great interference to lumen recognition and plaque analysis. In existing lumen segmentation techniques, the segmentation results of OCT images with blood are of low accuracy and poor robustness.

[0005] In the prior art, methods such as dynamic contour and region growing are often applied to the lumen segmentation of coronary OCT images, but these methods all have their own limitations. Specifically, when using the dynamic contour method for lumen segmentation, the initial lumen contour has a great influence on the final segmentation result, and this dynamic contour method cannot process OCT images with uncleaned blood in the coronary artery. When using the region growing method for lumen segmentation, the segmentation results of OCT images at bifurcations are often poor, and it also cannot process OCT images containing blood.

[0006] Moreover, existing methods such as dynamic contour and region growing often perform lumen segmentation on each cross-sectional image of coronary OCT images independently, which results in poor continuity of the finally segmented lumen contours. Summary of the Invention

[0007] In view of the above, an object of the present invention is to provide an automatic segmentation method, device, computing device, and storage medium for coronary OCT images, so as to improve the accuracy of lumen segmentation in OCT images containing blood and OCT images at bifurcations, and improve the continuity of lumen contours in adjacent frames.

[0008] In a first aspect, an automatic segmentation method for coronary OCT images provided by an embodiment includes the following steps:

[0009] Obtain an OCT image sequence obtained by performing OCT analysis on a target coronary blood vessel;

[0010] Coarse segmentation process: After converting the OCT image into a polar coordinate format image, calculate the boundary intensity value according to the gray value of the pixel point. According to the boundary intensity value of the pixel point, apply the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, and reflect the obtained lumen boundary onto the OCT image after smoothing to obtain a coarse segmentation result;

[0011] Fine segmentation process: Taking the lumen boundary points in the coarse segmentation result as the center, resample the OCT image along the outer normal direction, apply the dijksta minimum path algorithm to search for the lumen boundary in the resampled image, and reflect the obtained lumen boundary onto the OCT image after smoothing to obtain a fine segmentation result;

[0012] Map the lumen boundary in the fine segmentation result to the adjacent next frame of OCT image as the coarse segmentation result, and then perform the fine segmentation process to obtain the fine segmentation result.

[0013] In one embodiment, the conversion of the OCT image into a polar coordinate format image includes: constructing a polar coordinate system including an angle and a polar axis with the center of the OCT image as the pole, taking a plurality of discrete pixel points at intervals of length on the polar axis, taking a plurality of discrete pixel points at intervals of angle in the angle direction, and performing interpolation operations on each discrete pixel point to obtain a polar coordinate format image.

[0014] In one embodiment, the following formula is used to calculate the boundary intensity value according to the gray value of the pixel point:

[0015] Grad(i,j) = a * (I(i + 1,j) - I(i,j)) + b * ((I(i + 2,j) - I(i - 1,j)) + (I(i + 1,j - 1) - I(i,j - 1)) + (I(i + 1,j + 1) - I(i,j + 1))) + c * ((I(i + 2,j - 1) - I(i - 1,j - 1)) + (I(i + 2,j + 1) - I(i - 1,j + 1)))

[0016] Among them, Grad(i,j) represents the boundary intensity value of the pixel at position (i,j), I() represents the gray value of the pixel, and a, b, and c are distance weight coefficients respectively; the value range of the distance weight coefficient a is 0.8 - 1.2; the value range of the distance weight coefficient b is 0.4 - 0.6; the value range of the distance weight coefficient c is 0.2 - 0.3.

[0017] In one embodiment, when applying the dijksta minimum path algorithm to search for the lumen boundary in a polar coordinate format image, the reciprocal of the boundary intensity value is used as the consumption value of the pixel.

[0018] In one embodiment, before applying the dijksta minimum path algorithm to search for the lumen boundary in a polar coordinate format image, by setting a minimum value limit, the boundary intensity values less than the minimum limit value are normalized to the minimum limit value to achieve filtering preprocessing of the boundary intensity values; then, according to the filtered boundary intensity values, the dijksta minimum path algorithm is applied to search for the lumen boundary in the polar coordinate format image.

[0019] In one embodiment, when resampling the OCT image along the outer normal direction with the lumen boundary points in the rough segmentation result as the center, the value range of the sampling interval is set to 0.5 - 1.0 pixels.

[0020] In a second aspect, an automatic segmentation device for coronary OCT images provided by an embodiment includes:

[0021] An acquisition module, configured to acquire an OCT image sequence obtained by performing OCT analysis on a target coronary artery.

[0022] A rough segmentation module, configured to convert the OCT image into a polar coordinate format image, calculate the boundary intensity value according to the gray value of the pixel, apply the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image according to the boundary intensity value of the pixel, and reflect the obtained lumen boundary onto the OCT image after smoothing to obtain a rough segmentation result.

[0023] A fine segmentation module, which is used to resample the OCT image along the outer normal direction with the lumen boundary points in the rough segmentation result as the center, search for the lumen boundary in the resampled image by applying the dijksta minimum path algorithm, and reflect the obtained lumen boundary onto the OCT image after smoothing to obtain the fine segmentation result;

[0024] A transfer module, which is used to map the lumen boundary in the fine segmentation result to the adjacent next-frame OCT image as the rough segmentation result.

[0025] In a third aspect, a computing device provided by an embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the automatic segmentation method of the coronary OCT image described in the first aspect is implemented.

[0026] In a fourth aspect, a computer storage medium provided by an embodiment stores a computer program thereon. When the computer program is processed and executed, the automatic segmentation method of the coronary OCT image described in the first aspect is implemented.

[0027] The technical solutions provided by the above embodiments have at least the following beneficial effects:

[0028] The OCT image adopts two processes of rough segmentation and fine segmentation, and the fine segmentation result of the previous-frame OCT image is used as the rough segmentation result of the current-frame OCT image to continue the fine segmentation process, ensuring the continuity of the overall segmentation result of the lumen boundary in the OCT image sequence; in this way, for the OCT image with blood residue in the coronary lumen, based on the fine segmentation result of the previous OCT image without blood residue, fine segmentation is performed again, which can greatly reduce the influence of the residual blood on the lumen segmentation accuracy. Compared with the dynamic contour algorithm, the segmentation accuracy of the OCT image with blood residue can be significantly improved;

[0029] The dijkstra minimum path algorithm is used to search and identify the lumen boundary in the image. Compared with the region growing algorithm, it can accurately distinguish branch vessels and avoid inaccurate lumen boundary segmentation caused by overestimating the lumen at the bifurcation opening of the blood vessels. Description of the Drawings

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0031] Figure 1It is a flowchart of an automatic segmentation method for coronary OCT images provided by an embodiment;

[0032] Figure 2 It is the first-frame OCT image provided by an embodiment;

[0033] Figure 3 It is the polar coordinate format image converted from the OCT image provided by an embodiment;

[0034] Figure 4 It is the lumen boundary identified by performing the dijkstra minimum path algorithm search in the polar coordinate format image provided by an embodiment;

[0035] Figure 5 It is the smoothed result graph of the lumen boundary provided by an embodiment;

[0036] Figure 6 It is a schematic diagram of mapping the smoothed lumen boundary to the first-frame OCT image provided by an embodiment;

[0037] Figure 7 It is the original tenth-frame OCT image provided by an embodiment;

[0038] Figure 8 It is the resampled image corresponding to the tenth-frame OCT image provided by an embodiment;

[0039] Figure 9 It is the lumen boundary identified by performing the dijkstra minimum path algorithm search in the resampled image provided by an embodiment;

[0040] Figure 10 It is a schematic diagram of mapping the lumen boundary to the tenth-frame OCT image provided by an embodiment;

[0041] Figure 11 It is a schematic diagram of comparing the lumen boundaries between the ninth image and the tenth image provided by an embodiment;

[0042] Figure 12 It is a schematic diagram of the segmentation result of the bifurcated OCT image provided by an embodiment;

[0043] Figure 13 It is a schematic diagram of the segmentation result of the OCT image with blood residue provided by an embodiment;

[0044] Figure 14 It is a schematic diagram of the structure of an automatic segmentation device for coronary OCT images provided by an embodiment. Detailed implementation manners

[0045] To make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and do not limit the protection scope of the present invention.

[0046] In view of the problem that the existing technology has inaccurate lumen segmentation for OCT images with blood residue and OCT images at bifurcations, the embodiment provides an automatic segmentation method and device for coronary OCT images.

[0047] Figure 1 is a flowchart of an automatic segmentation method for coronary OCT images provided by an embodiment. As Figure 1 shown, the automatic segmentation method for coronary OCT images provided by the embodiment includes the following steps:

[0048] S101, obtaining an OCT image sequence obtained by performing OCT analysis on a target coronary artery.

[0049] In the embodiment, an OCT image sequence is obtained by performing OCT analysis on a target coronary artery. This OCT image sequence is essentially a continuous cross-sectional image of a target coronary artery. Therefore, the coronary OCT image sequence is composed of N consecutive cross-sectional OCT images, which are respectively denoted as I 1 、I 2 、I 3 …I N . For each frame of OCT image, two processes of rough segmentation and fine segmentation are performed to obtain the lumen boundary contour in the OCT image.

[0050] S102, performing rough segmentation on the first frame of OCT image in the OCT image sequence.

[0051] In the embodiment, the process of performing rough segmentation on the first frame of OCT image includes: after converting the OCT image into a polar coordinate format image, calculating the boundary intensity value according to the gray value of the pixel point, and according to the boundary intensity value of the pixel point, applying the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, and reflecting the obtained lumen boundary onto the OCT image after smoothing to obtain the rough segmentation result.

[0052] Figure 2 is an exemplary illustration of the first frame of OCT image. Figure 2The coordinate system of the first-frame OCT image shown is a rectangular coordinate system, which can coincide with the real cross-section of the blood vessel. Based on this, in a possible implementation, converting the OCT image into a polar coordinate format image includes: constructing a polar coordinate system with the center of the OCT image as the pole, which contains an angle and a polar axis, taking multiple discrete pixel points at length intervals on the polar axis, taking multiple discrete pixel points at angle intervals in the angle direction, and performing interpolation operations on each discrete pixel point to obtain the polar coordinate format image.

[0053] Exemplarily, the origin of the rectangular coordinate system of the OCT image, that is, the center of the OCT image, can be used as the pole, and the mutually perpendicular x-axis and y-axis can be used as the polar axis (or 0° direction) and 90° direction of the polar coordinate system respectively. In the polar axis direction, n discrete points are taken at a length interval of spa 1 discrete points are taken at an angle interval of rad in the angle direction, preferably, the size of spa is 1 - 4 pixel units, and n 2 the size of n is 50–400, the size of rad is 1°-9°, and the size of n 1 is 40–360; then bilinear interpolation or other interpolation methods are used to perform interpolation calculations on each discrete point to obtain the new pixel value corresponding to the discrete point, and thus the polar coordinate format image is obtained. For the 2 shown OCT image, conversion is performed according to spa = 1.5, n Figure 2 = 200, rad = 3.6°, n 1 = 100 to obtain the 2 shown polar coordinate format image. Figure 3 shown polar coordinate format image.

[0054] After obtaining the polar coordinate format image with a size of n 1 *n 2 For the pixel point at any position (i, j), its gray value is I(i, j), and the boundary strength is Grad(i, j). In general processing methods such as the Roberts operator and the Sobel operator, the boundary strength is defined as the absolute value of the gray gradient. However, in the present invention, the coronary OCT image shows the characteristics of low gray values in the lumen area and high gray values in the vascular wall area. For this image, directly defining the absolute value of the gray gradient as the boundary strength is not accurate. Through experimental exploration, in the embodiment, the boundary strength is defined as the gray gradient in the polar axis direction, that is, the boundary strength value is calculated according to the gray value of the pixel point by using the following formula:

[0055] Grad(i,j) = a * (I(i + 1,j) - I(i,j)) + b * ((I(i + 2,j) - I(i - 1,j)) + (I(i + 1,j - 1) - I(i,j - 1)) + (I(i + 1,j + 1) - I(i,j + 1))) + c * ((I(i + 2,j - 1) - I(i - 1,j - 1)) + (I(i + 2,j + 1) - I(i - 1,j + 1)))

[0056] Among them, Grad(i,j) represents the boundary intensity value of the pixel at position (i,j), I() represents the gray value of the pixel, and a, b, and c are distance weight coefficients respectively. Preferably, the value range of a is 0.8 - 1.2; the value range of b is 0.4 - 0.6; the value range of c is 0.2 - 0.3.

[0057] It can be seen from the above formula for calculating the boundary intensity value that the boundary intensity of any pixel is determined by its adjacent 3 * 4 points. The larger the boundary intensity value, the greater the possibility that it is the lumen boundary, and the smaller the boundary intensity value, the smaller the possibility that it is the lumen boundary.

[0058] In the embodiment, when searching for the lumen boundary in the polar coordinate format image according to the boundary intensity value of the pixel by applying the dijksta minimum path algorithm, the reciprocal of the boundary intensity value is used as the consumption value cost of the pixel, that is, cost = 1 / Grad.

[0059] It should be noted that it can be seen from the formula for calculating the boundary intensity value that for some pixels, their boundary intensity values may be negative. Negative boundary intensity values and too small negative boundary intensity values cannot be well applied to the dijksta minimum path algorithm for lumen boundary search. Based on this, in the embodiment, before applying the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, by setting the minimum limit value σ of the boundary intensity value, the boundary intensity value less than the minimum limit value is normalized to the minimum limit value, that is, if Grad < σ, then let Grad = σ, to achieve the filtering preprocessing of the boundary intensity value; then, according to the filtered and preprocessed boundary intensity value, apply the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, and the obtained lumen boundary is as Figure 4 shown. In the embodiment, the minimum limit value σ of the boundary intensity value can be set to 5 - 10.

[0060] After obtaining the lumen boundary, perform a certain degree of smoothing processing on the obtained lumen boundary. For Figure 4 the lumen boundary shown, the result of the smoothing processing is as Figure 5 shown, and then Figure 5 the smoothed lumen boundary shown is mapped to the OCT image, and the result is as Figure 6The rough segmentation result shown

[0061] S103, perform fine segmentation on the OCT image, and use the obtained fine segmentation result as the segmentation result of the OCT image.

[0062] In the embodiment, the process of performing fine segmentation on the rough segmentation result is as follows: taking the lumen boundary points in the rough segmentation result as the center, resampling the OCT image along the outer normal direction, applying the dijksta minimum path algorithm to search for the lumen boundary in the resampled image, and reflecting the obtained lumen boundary back to the OCT image after smoothing to obtain the fine segmentation result.

[0063] When resampling the rough segmentation result, set the value range of the sampling interval to 0.5 - 1.0 pixel units. Take each lumen boundary point as the center one by one, and perform resampling along the outer normal direction according to the sampling interval to obtain the resampled image. Then apply the dijksta minimum path algorithm to the resampled image to identify a finer lumen boundary. This finer lumen boundary is mapped back to the original first-frame OCT image after smoothing filtering, that is, the first-frame OCT image I 1 The final segmentation result of the segmentation.

[0064] S104, map the lumen boundary in the fine segmentation result to the adjacent next-frame OCT image as the rough segmentation result, and then perform the fine segmentation process.

[0065] In the OCT image sequence, the spatial interval between two adjacent OCT images is small. Therefore, there will be no large mutations in the lumen boundaries corresponding to two adjacent OCT images. After obtaining the final lumen boundary of the fine segmentation of the OCT image I 1 map the final lumen boundary to the OCT image I 2 as the rough segmentation result of the OCT image I 2 Apply the same method as the fine segmentation of the OCT image I 1 to obtain the fine segmentation lumen boundary of the OCT image I 2 Perform iterative segmentation according to this method until the segmentation of the last OCT image I N ends, and the final segmentation result of the entire coronary artery OCT image sequence can be obtained for subsequent analysis and processing of morphological parameters such as the minimum lumen diameter, minimum lumen area, and stenosis degree.

[0066] For the tenth-frame OCT image I Figure 7 shown as 10 map the final lumen boundary of the ninth-frame OCT image I 9 to the tenth-frame OCT image as the rough segmentation result of the tenth-frame OCT image I 10 and perform fine segmentation on the tenth-frame OCT image I 10The rough segmentation result is resampled at a sampling interval of Δs = 0.8 to obtain a resampled image as shown in Figure 8 . Then, the Dijkstra minimum path algorithm is used to search for the lumen boundary in the resampled image, and the obtained lumen boundary is mapped to the tenth-frame OCT image I Figure 9 to obtain a fine segmentation result as shown in 10 . The lumen boundary in the fine segmentation result is compared with the lumen boundary in the ninth-frame OCT image I Figure 10 as shown in 9 . Figure 11

[0067] The segmentation results of the bifurcated OCT image and the OCT image with blood residue using the above automatic segmentation method for coronary OCT images are shown in Figure 12 and Figure 13 . It can be seen from the analysis of Figure 12 and Figure 13 that the lumen boundary is clearly and accurately presented in the segmentation results.

[0068] Figure 14 is a schematic structural diagram of an automatic segmentation device for coronary OCT images provided by an embodiment. As shown in Figure 14 , the automatic segmentation device 1400 provided by the embodiment includes:

[0069] An acquisition module 1401, configured to acquire an OCT image sequence obtained by performing OCT analysis on a target coronary blood vessel;

[0070] A rough segmentation module 1402, configured to convert the OCT image into a polar coordinate format image, calculate the boundary intensity value according to the gray value of the pixel points, and apply the Dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, and map the obtained lumen boundary to the OCT image after smoothing to obtain a rough segmentation result;

[0071] A fine segmentation module 1403, configured to resample the OCT image along the outer normal direction with the lumen boundary points in the rough segmentation result as the center, apply the Dijksta minimum path algorithm to search for the lumen boundary in the resampled image, and map the obtained lumen boundary to the OCT image after smoothing to obtain a fine segmentation result;

[0072] A transfer module 1404, configured to map the lumen boundary in the fine segmentation result to the adjacent next-frame OCT image as the rough segmentation result.

[0073] ​It should be noted that when the automatic segmentation device for coronary OCT images provided in the above embodiments performs lumen boundary segmentation on coronary OCT images, the above-described functional modules should be used as examples for illustration. The above functions can be allocated to different functional modules according to needs, that is, the internal structure of the terminal or server is divided into different functional modules to complete all or part of the functions described above. In addition, the automatic segmentation device for coronary OCT images provided in the above embodiments and the embodiments of the automatic segmentation method for coronary OCT images belong to the same concept. For the specific implementation process, please refer to the embodiments of the automatic segmentation method for coronary OCT images, which will not be elaborated here.

[0074] For the above automatic segmentation method and device for coronary OCT images, the OCT images are subjected to a rough segmentation process and a fine segmentation process, and the fine segmentation result of the previous frame of OCT image is used as the rough segmentation result of the current frame of OCT image, and the fine segmentation process is continued, which ensures the continuity of the overall segmentation result of the lumen boundary in the OCT image sequence. In this way, for OCT images with blood residue in the coronary lumen, based on the fine segmentation results of the previous OCT images without blood residue, fine segmentation can be performed again, which can greatly reduce the influence of residual blood on the lumen segmentation accuracy. Compared with the dynamic contour algorithm, the segmentation accuracy of OCT images with blood can be significantly improved.

[0075] The dijkstra minimum path algorithm is used to search and identify the lumen boundary in the image. Compared with the region growing algorithm, it can accurately distinguish branch vessels and avoid inaccurate lumen boundary segmentation caused by overestimating the lumen at the bifurcation opening of blood vessels.

[0076] The embodiment also provides a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above automatic segmentation method for coronary OCT images is implemented, that is, the following steps are implemented:

[0077] S101, obtaining an OCT image sequence obtained by performing OCT analysis on a target coronary blood vessel;

[0078] S102, performing rough segmentation on the first frame of OCT image in the OCT image sequence;

[0079] S103, performing fine segmentation on the rough segmentation result, and using the obtained fine segmentation result as the segmentation result of the OCT image;

[0080] S104, mapping the lumen boundary in the fine segmentation result to the adjacent next frame of OCT image as the rough segmentation result, and then performing the fine segmentation process.

[0081] In practical applications, the memory can be a volatile memory proximal to the device, such as RAM, or a non-volatile memory, such as ROM, FLASH, floppy disks, mechanical hard disks, etc., or it can also be a remote storage cloud. The processor can be a central processing unit (CPU), a microprocessor (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA). That is, the automatic segmentation steps of coronary OCT images can be implemented through these processors.

[0082] The embodiment also provides a computer storage medium, on which a computer program is stored. When the computer program is processed and executed, it implements the above-mentioned automatic segmentation method of coronary OCT images, that is, the following steps are implemented:

[0083] S101, obtaining an OCT image sequence obtained by performing OCT analysis on a target coronary artery;

[0084] S102, performing rough segmentation on the first frame of the OCT image sequence;

[0085] S103, performing fine segmentation on the rough segmentation result, and using the obtained fine segmentation result as the segmentation result of the OCT image;

[0086] S104, mapping the lumen boundary in the fine segmentation result to the adjacent next frame of the OCT image as the rough segmentation result, and then performing the fine segmentation process.

[0087] In practical applications, the computer-readable storage medium can be ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0088] The above specific embodiments have described in detail the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, supplements, equivalent replacements, etc. made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An automatic segmentation method for coronary OCT images, characterized in that, it includes the following steps: Obtain an OCT image sequence obtained by performing OCT analysis on a target coronary blood vessel; Coarse segmentation process: After converting the OCT image into a polar coordinate format image, define the boundary intensity as the gray gradient in the polar axis direction, calculate the boundary intensity value based on the gray value of the pixel point, and the boundary intensity value of any pixel point is determined by the adjacent 3*4 points of the corresponding coordinate point. By setting a minimum limit, the boundary intensity values less than the minimum limit value are normalized to the minimum limit value to achieve filtering preprocessing of the boundary intensity values. According to the filtered boundary intensity values, apply the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, and reflect the obtained lumen boundary to the OCT image after smoothing to obtain the coarse segmentation result; Use the following formula to calculate the boundary intensity value based on the gray value of the pixel point: Grad(i,j)=a*(I(i+1,j)-I(i,j))+b*((I(i+2,j)- I(i-1,j))+(I(i+1,j-1)- I(i,j-1))+ (I(i+1,j+1)- I(i,j+1)))+c*((I(i+2,j-1)- I(i-1,j-1))+(I(i+2,j+1)- I(i-1,j+1))) where, Grad(i,j) represents the boundary intensity value of the pixel point at position (i,j), I() represents the gray value of the pixel point, and a, b, and c are distance weight coefficients respectively; Fine segmentation process: Take the lumen boundary points in the coarse segmentation result as the center, resample the OCT image along the outer normal direction, apply the dijksta minimum path algorithm to search for the lumen boundary in the resampled image, and reflect the obtained lumen boundary to the OCT image after smoothing to obtain the fine segmentation result; Map the lumen boundary in the fine segmentation result to the adjacent next frame of OCT image as the coarse segmentation result, and then perform the fine segmentation process to obtain the fine segmentation result.

2. The automatic segmentation method for coronary OCT images according to claim 1, characterized in that, the conversion of the OCT image into a polar coordinate format image includes: constructing a polar coordinate system including an angle and a polar axis with the center of the OCT image as the pole, taking a plurality of discrete pixel points at equal length intervals on the polar axis, taking a plurality of discrete pixel points at equal angle intervals in the angle direction, and performing interpolation operations on each discrete pixel point to obtain a polar coordinate format image.

3. The automatic segmentation method for coronary OCT images according to claim 1, characterized in that, the value range of the distance weight coefficient a is 0.8 - 1.2; the value range of the distance weight coefficient b is 0.4 – 0.6; the value range of the distance weight coefficient c is 0.2 – 0.

3.

4. The automatic segmentation method for coronary OCT images according to claim 1, characterized in that, when applying the dijksta minimum path algorithm to search for the lumen boundary in the polar coordinate format image, use the reciprocal of the boundary intensity value as the consumption value of the pixel point.

5. The automatic segmentation method of coronary OCT images according to claim 1, characterized in that, when resampling the OCT image along the outer normal direction with the lumen boundary points in the rough segmentation result as the center, the value range of the sampling interval is set to 0.5 - 1.0 pixels.

6. An automatic segmentation device for coronary OCT images, characterized in that, comprising: an acquisition module for acquiring an OCT image sequence obtained by performing OCT analysis on a target coronary artery; a rough segmentation module for converting the OCT image into a polar coordinate format image, defining the boundary intensity as the gray gradient in the polar axis direction, calculating the boundary intensity value according to the gray value of the pixel point, the boundary intensity value of any pixel point being determined by the adjacent 3*4 points of the corresponding coordinate point, and normalizing the boundary intensity value less than the minimum limit value to the minimum limit value by setting the minimum limit value to achieve filtering preprocessing of the boundary intensity value, and searching for the lumen boundary in the polar coordinate format image by applying the dijksta minimum path algorithm, and reflecting the obtained lumen boundary onto the OCT image after smoothing to obtain a rough segmentation result; The boundary intensity value is calculated according to the gray value of the pixel point by using the following formula: Grad(i,j)=a*(I(i+1,j)-I(i,j))+b*((I(i+2,j)- I(i-1,j))+(I(i+1,j-1)- I(i,j-1))+ (I(i+1,j+1)- I(i,j+1)))+c*((I(i+2,j-1)- I(i-1,j-1))+(I(i+2,j+1)- I(i-1,j+1))) where Grad(i,j) represents the boundary intensity value of the pixel point at position (i,j), I() represents the gray value of the pixel point, and a, b, and c are distance weight coefficients respectively; a fine segmentation module for resampling the OCT image along the outer normal direction with the lumen boundary points in the rough segmentation result as the center, searching for the lumen boundary in the resampled image by applying the dijksta minimum path algorithm, and reflecting the obtained lumen boundary onto the OCT image after smoothing to obtain a fine segmentation result; a transfer module for mapping the lumen boundary in the fine segmentation result to the adjacent next frame of OCT image as the rough segmentation result.

7. A computing device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, when the processor executes the computer program, the automatic segmentation method of coronary OCT images according to any one of claims 1 to 5 is implemented.

8. A computer storage medium, on which a computer program is stored, characterized in that, when the computer program is processed and executed, the automatic segmentation method of coronary OCT images according to any one of claims 1 to 5 is implemented.

Images