Computer vision-based lung region segmentation system for chronic obstructive pulmonary disease
By combining a segmentation system that reflects the brightness characteristics of the lung area, the segmentation of lung and non-pulmonary areas and morphological processing, the problems of inaccurate lung image segmentation and lack of adaptability in the prior art are solved, and high-precision and flexible segmentation of lung area are achieved.
Patent Information
- Application Number
- CN202510765219.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-10
- Publication Date
- 2025-07-08
AI Technical Summary
The existing medical imaging segmentation methods are difficult to adapt to lung images in different patients and scan conditions, and are difficult to effectively remove noise, resulting in inaccurate segmentation results and lack of adaptability.
Using a segmentation system based on computer vision, the segmentation results are dynamically adjusted by reflecting the combination of the pulmonary region brightness characteristic unit, the segmentation unit of lung and non-pulmonary region, and the morphological processing and segmentation result optimization unit.
The precise segmentation of lung images is achieved, the anti-noise ability is enhanced, the flexibility and accuracy of the segmentation system are improved, and the lung images are adapted to different patients and scanned conditions.
Smart Images
Figure CN120279048A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical image processing, and particularly to a system for segmenting the lung region of chronic obstructive pulmonary disease (COPD) based on computer vision. Background Art
[0002] Chronic obstructive pulmonary disease (COPD) is a common respiratory disease. Its early diagnosis and treatment are crucial for improving the quality of life of patients. Computer vision is an important branch of artificial intelligence, which studies how to extract useful information from images and videos, and understand, analyze, and process them. With the continuous progress of computer vision technology, its application in the field of medical image processing is becoming increasingly widespread. Specifically, through computer vision technology, automatic analysis, diagnosis, and auxiliary decision-making of medical images can be achieved, thereby improving the efficiency and quality of medical services.
[0003] However, existing medical image segmentation methods often rely on fixed thresholds and simple morphological operations, and thus are difficult to adapt to lung images of different patients and under different scanning conditions. This results in errors in the segmentation results, affecting subsequent diagnosis and treatment. There are often a large number of noises and artifacts in the lung CT images, and these noises will interfere with the segmentation results. Existing segmentation methods often have difficulty effectively removing these noises, leading to inaccurate segmentation results. In addition, existing segmentation methods often lack adaptability and cannot be dynamically adjusted according to the quality of the segmentation results, which limits the flexibility and accuracy of the segmentation system. Summary of the Invention
[0004] The purpose of the present invention is to provide a system for segmenting the lung region of chronic obstructive pulmonary disease (COPD) based on computer vision, which solves the problems raised in the above background art.
[0005] To achieve the above purpose, the present invention provides the following technical solutions, and the specific implementation steps are as follows: Step 1: Use an image data acquisition module to obtain the original lung CT image of the patient; Step 2: Use an input module to input the lung CT image into a segmentation processing module; Step 3: The segmentation processing module sequentially calculates and outputs the gray value H of the i-th pixel point i 、the segmentation result FG of the i-th pixel point i 、the final segmentation result FGJ; Step 4: Present the final segmentation optimization result of the final segmentation result FGJ on a display module; Among them, the segmentation processing module includes a unit for reflecting the brightness characteristics of the lung region, a unit for segmenting the lung and non-lung regions, and a unit for morphological processing and optimizing the segmentation result; When acquiring the original lung CT image of the patient, the gray value H of the i-th pixel point needs to be calculated one by one according to each preset pixel area of the lung i , and based on the gray value H of the i-th pixel point i , judge the segmentation result FG of the i-th pixel point in each pixel area i , and then gather all the segmentation results FG of the i-th pixel point i , after performing dilation and erosion processing, calculate the final segmentation result FGJ; The devices used by the image data acquisition module include a CT scanner; The devices used by the input module include an input device; The devices used by the segmentation processing module include a computer and image processing software; The devices used by the display module include a display.
[0006] Optionally, the calculation formula of the unit reflecting the brightness characteristics of the lung area is as follows: ; Where: H i is the gray value of the i-th pixel point; R i is the red channel value of the i-th pixel point, reflecting the red component in the image; G i is the green channel value of the i-th pixel point, reflecting the green component in the image, and plays a key role in identifying the structure and texture of the lung in the lung CT image; B i is the blue channel value of the i-th pixel point, reflecting the green component in the image; is the red channel value R of the i-th pixel point i , the green channel value G of the i-th pixel point i and the blue channel value B of the i-th pixel point i each represent different color components and information in the color image, and when processing the lung CT image, the information of the above three channels jointly determines the color and brightness of the image.
[0007] Optionally, the calculation formula of the lung and non-lung area segmentation unit is as follows: FG i =(H i -FGH i )×sgn(H i -FGH i )+a×H avg ; ; Where: FGi is the segmentation result of the i-th pixel, and its value range is {0 - 1}. If it is determined to be 1, it is the lung region; if it is determined to be 0, it is the non-lung region; FGH i is the segmentation threshold of the i-th pixel; a is the neighborhood influence coefficient, and its value range is {0 - 1}, and it is used to adjust the influence of neighborhood pixels on the segmentation result: When a = 0, it means not considering the neighborhood influence; When a = 1, it means completely relying on neighborhood pixels for segmentation; H avg is the average gray value, reflecting the average degree of gray values within the neighborhood of the current pixel; N is the total number of pixels, reflecting the total number of all pixels within the neighborhood of the current pixel; sgn(H i - FGH i ) is specifically the sign function; When H i - FGH i > 0, sgn(H i - FGH i ) = 1; When H i - FGH i ≤ 0, sgn(H i - FGH i ) = -1.
[0008] Optionally, the non-consideration of neighborhood influence is as follows: When the image quality is high and the contrast between the lung region and the non-lung region is obvious, the gray value of a single pixel is sufficient for accurate segmentation. At this time, the neighborhood influence is not considered, that is, set a = 0; When the processing speed is a key factor and the accuracy requirement for the segmentation result is not high, the neighborhood influence is not considered, that is, set a = 0; Completely relying on neighborhood pixels for segmentation is as follows: When the image has a lot of noise and the contrast between the lung region and the non-lung region is low, the gray value of a single pixel is not sufficient for accurate segmentation. At this time, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, set a = 1; When processing images with complex textures and structures, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, set a = 1; Among them, according to the set situation, the system automatically adjusts the value of the neighborhood influence coefficient a.
[0009] Optionally, the system calculates the gray value H of the i-th pixel for all pixels of the patient's lungs iand the segmentation result FGi of the i-th pixel i Perform one-to-one calculation and output, and perform dilation and erosion operations, specifically as follows: ; ; Where: FG p is the result after dilation, and it is the image after performing the dilation operation on the segmentation result FGi of the i-th pixel i ; FG F is the result after erosion, and it is the image after performing the erosion operation on the segmentation result FGi of the i-th pixel i ; ⊕ represents the dilation operation, represents the erosion operation; YS is the structuring element, and it reflects the pixel area of a small rectangle and a circle.
[0010] Optionally, the calculation formula of the morphological processing and segmentation result optimization unit is as follows: FGJ = (FG p - FG F ) × β + FG F ; Where: FGJ is the final segmentation result; β is the weight coefficient of the result FG p after dilation and the result FG F after erosion, and its value range is {0 - 1}, which is used to adjust the influence of dilation and erosion on the final result; The final segmentation result FGJ is used as the input value for subsequent processing, including feature extraction and quality assessment.
[0011] Optionally, based on the weight coefficient β When the weight coefficient β increases, that is, approaches 1, the contribution of the dilation result to the final segmentation result increases, resulting in an expansion of the lung region; When the weight coefficient β decreases, that is, approaches 0, the contribution of the erosion result to the final segmentation result increases, resulting in a shrinkage of the lung region.
[0012] Optionally, the calculation formula of the segmentation threshold FGH of the i-th pixel i is as follows: ; M is the total number of patients; Among them, according to the segmentation result FGi of the i-th pixel of the total number of patients M calculated recently i , collect and calculate, and calculate and output the segmentation threshold FGH of the i-th pixel iA reference value that is updated regularly In addition, the segmentation threshold FGH of the i-th pixel point needs to be calculated for each pixel point in the pixel area of the lung image i For reference
[0013] Compared with the prior art, the beneficial effects of the present invention are as follows 1. The present invention realizes the accurate segmentation of lung images through the combination of the unit reflecting the brightness characteristics of the lung region and the unit for segmenting the lung and non-lung regions. Among them, the unit reflecting the brightness characteristics of the lung region provides an accurate basis for subsequent segmentation by calculating the gray value, while the unit for segmenting the lung and non-lung regions uses the threshold segmentation of each pixel point and the neighborhood information to effectively reduce the segmentation error and thus improve the segmentation accuracy
[0014] 2. The present invention effectively enhances the anti-noise ability of the system through the neighborhood influence coefficient a in the unit for segmenting the lung and non-lung regions and the morphological processing in the unit for optimizing the morphological processing and segmentation results. Among them, the neighborhood influence coefficient a can smooth the segmentation edge, thereby reducing the influence of noise on the segmentation result, while the morphological processing can further remove small targets and fill holes to improve the stability and accuracy of the segmentation result
[0015] 3. The present invention realizes the adaptive adjustment of the segmentation result through the weight coefficient β in the unit for optimizing the morphological processing and segmentation results, and can dynamically adjust the value of β according to the quality of the segmentation result, thereby affecting the segmentation threshold FGH of the i-th pixel point in the unit for segmenting the lung and non-lung regions i And the setting of the neighborhood influence coefficient a. This cyclic influence mechanism makes the entire segmentation system adaptive, capable of dynamically adjusting according to specific situations, and improving the flexibility and accuracy of the segmentation system
[0016] The algorithm formula proposed by this system not only considers the gray value and threshold segmentation of the image, but also introduces neighborhood information and morphological processing to realize the accurate segmentation of lung images. At the same time, through the cyclic influence mechanism, the adaptive adjustment of the segmentation result is realized, and the flexibility and accuracy of the system are improved Brief Description of the Drawings
[0017] Figure 1 is the method flow chart of the chronic obstructive pulmonary disease lung region segmentation system based on computer vision Figure 2 is the structural schematic diagram of the segmentation processing module of the present invention Figure 3 is the schematic diagram of the determination result of the segmentation result of the present invention Detailed Embodiments
[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0019] Regarding this computer vision-based COPD lung region segmentation system, different from the existing medical image segmentation systems, the existing medical image segmentation systems are difficult to adapt to lung images of different patients and different scanning conditions, and are difficult to remove the influence of noise, and also lack self-adaptability. However, this algorithm unit improves the segmentation accuracy, enhances the anti-noise ability, and realizes adaptive segmentation. And these beneficial effects make the system have important clinical significance and practical application value in the early diagnosis and treatment of COPD.
[0020] Example 1, please refer to Figures 1 to 3 , this embodiment provides a computer vision-based COPD lung region segmentation system, and the specific implementation steps are as follows: Step 1: Use the image data acquisition module to obtain the original lung CT image of the patient; Step 2: Use the input module to input the lung CT image into the segmentation processing module; Step 3: The segmentation processing module sequentially calculates and outputs the gray value H of the i-th pixel point i , the segmentation result FG of the i-th pixel point i , and the final segmentation result FGJ; Step 4: Present the final segmentation optimization result of the final segmentation result FGJ on the display module; Among them, the segmentation processing module includes a unit reflecting the brightness characteristics of the lung region, a unit for segmenting the lung and non-lung regions, and a unit for morphological processing and optimizing the segmentation result; When obtaining the original lung CT image of the patient, it is necessary to calculate and output the gray value H of the i-th pixel point one by one according to each preset pixel region of the lung i , and based on the gray value H of the i-th pixel point i , judge the segmentation result FG of the i-th pixel point in each pixel region i , and then combine all the segmentation results FG of the i-th pixel point i , perform dilation and erosion processing and then calculate the final segmentation result FGJ; The devices used in the image data acquisition module include a CT scanner; The devices used in the input module include an input device; The devices used in the segmentation processing module include a computer and image processing software; The devices used by the display module include a display.
[0021] In this embodiment, the system cooperates with three algorithm units and combines H i , FG i and the three operation results of FGJ to jointly constitute the core algorithm of the COPD lung region segmentation system based on computer vision. Specifically, H i is the gray value of the i-th pixel. After improving the image quality through preprocessing, calculating the gray value can more accurately reflect the brightness characteristics of the lung region and provide a reliable input for the subsequent segmentation algorithm. FG i is the segmentation result of the i-th pixel. The calculation purpose also refers to using the threshold segmentation method to segment the lung region from the gray image. This step is the core part of the COPD lung region segmentation system because accurate segmentation results are crucial for subsequent feature extraction and quality assessment processing. FGJ is the final segmentation result. By performing dilation and erosion operations on FG i , two images after dilation and erosion can be obtained. Then, these two images are added together according to the weight to obtain the final FGJ. This step can remove small targets and holes in the preliminary segmentation result, and at the same time smooth the segmentation edge, making the segmentation result more complete and accurate. Moreover, the calculation result of FGJ can also affect the feedback to H i and FG i calculations, making the three algorithms of the system each have significant beneficial effects. The cyclic influence of FGJ on H i and FG i further enhances the self-adaptability and optimization ability of the entire segmentation system. These beneficial effects act together on the COPD lung region segmentation system based on computer vision, improving the accuracy, stability, and efficiency of segmentation.
[0022] Please refer to Figures 1 to 3 , the calculation formula of the unit reflecting the brightness characteristics of the lung region is as follows: ; Among them: H i is the gray value of the i-th pixel; R i is the red channel value of the i-th pixel, reflecting the red component in the image; G i is the green channel value of the i-th pixel, reflecting the green component in the image and playing a key role in identifying the structure and texture of the lungs in lung CT images; B i is the blue channel value of the i-th pixel, reflecting the green component in the image; is the red channel value R of the i-th pixeli The green channel value G of the i-th pixel i and the blue channel value B of the i-th pixel i each represent different color components and information in the color image. When processing lung CT images, the information of these three channels jointly determines the color and brightness of the image.
[0023] In this embodiment: First, the " " calculation part in this algorithm unit is to extract grayscale information from the original RGB color image. The grayscale value is a basic attribute of the image, which reflects the brightness information of the image. In image processing, grayscale images are easier to process than color images because grayscale images only contain information of one channel. By performing a sum-of-squares operation on the RGB three-channel values, the contribution of the three channels to brightness can be comprehensively considered, thereby obtaining a more accurate grayscale value. This calculation result is the grayscale value H of the i-th pixel in the unit reflecting the brightness characteristics of the lung region i , that is, the grayscale value of the preprocessed image. This grayscale value will be used as the input for the subsequent segmentation algorithm to determine whether each pixel belongs to the lung region; This step of converting the RGB three-channel image to a grayscale image in this algorithm unit not only simplifies the subsequent processing flow but also significantly reduces the computational complexity. Grayscale images only contain brightness information, making the subsequent processing more efficient. Among them, the grayscale value reflects the brightness information of the image. For lung CT images, the brightness difference between the lung region and other tissues is the key to segmentation. The unit reflecting the brightness characteristics of the lung region provides an accurate brightness information basis for the subsequent segmentation processing by precisely calculating the grayscale value, thereby improving the accuracy of segmentation; In practical applications, lung CT images are often affected by noise and illumination changes. The unit reflecting the brightness characteristics of the lung region effectively reduces the influence of these interference factors on the image by performing a square root operation on the sum of squares of the RGB three-channel values, enhancing the robustness of the algorithm. This operation method makes the grayscale value have a certain robustness to noise and illumination changes, thereby improving the stability of the segmentation result.
[0024] Please refer to Figures 1 to 3 , the calculation formula of the lung and non-lung region segmentation unit is as follows: FG i =(H i -FGH i )×sgn(H i -FGH i )+a×H avg ; ; Where: FG iis the segmentation result of the i-th pixel, and its value range is {0-1}. If it is determined to be 1, it is the lung region; if it is determined to be 0, it is the non-lung region; FGH i is the segmentation threshold of the i-th pixel; a is the neighborhood influence coefficient, and its value range is {0-1}, and it is used to adjust the influence of neighborhood pixels on the segmentation result: When a = 0, it means that the neighborhood influence is not considered; When a = 1, it means that the segmentation is completely dependent on neighborhood pixels; H avg is the average gray value, which reflects the average degree of gray values within the neighborhood of the current pixel; N is the total number of pixels, which reflects the total number of all pixels within the neighborhood of the current pixel; sgn(H i -FGH i ) is specifically the sign function; When H i -FGH i > 0, sgn(H i -FGH i ) = 1; When H i -FGH i ≤ 0, sgn(H i -FGH i ) = -1; The non-consideration of neighborhood influence is as follows: When the image quality is high and the contrast between the lung region and the non-lung region is obvious, the gray value of a single pixel is sufficient for accurate segmentation. At this time, the neighborhood influence is not considered, that is, a = 0 is set; When the processing speed is a key factor and the accuracy requirement for the segmentation result is not high, the neighborhood influence is not considered, that is, a = 0 is set; The complete dependence on neighborhood pixels for segmentation is as follows: When the image noise is large and the contrast between the lung region and the non-lung region is low, the gray value of a single pixel is not sufficient for accurate segmentation. At this time, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, a = 1 is set; When processing images with complex textures and structures, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, a = 1 is set; Among them, according to the set situation, the system automatically adjusts the value of the neighborhood influence coefficient a.
[0025] In this embodiment, first, the "(Hi - FGHi)" calculation part is used to compare the gray value H i of the i-th pixel in the preprocessed image with the segmentation threshold FGH of the i-th pixeli , where the threshold segmentation is a simple and effective image segmentation method. It is based on whether the gray value H of the i-th pixel point i exceeds the segmentation threshold FGH of the i-th pixel point i to segment the image into different regions. In this formula, by calculating "(H i - FGH i ), it is possible to determine whether each pixel point belongs to the lung region. And this calculation result is the input of "sgn(H i - FGH i )" for determining whether each pixel point belongs to the lung region. Specifically, the calculation part of "sgn(H i - FGH i )" is used to classify pixel points into two categories according to the result of "(H i - FGH i ). One is that it belongs to the lung region where sgn(H i - FGH i ) = 1, and the other is that it does not belong to the lung region where sgn(H i - FGH i ) = -1. This function plays a key role in threshold segmentation. It binarizes the image according to the comparison result of the gray value and the threshold, and then obtains a preliminary segmentation result; The calculation part of "a×H avg " is to introduce neighborhood information to smooth the segmentation edge and reduce the influence of noise on the segmentation result. a is the neighborhood influence coefficient used to adjust the influence degree of neighborhood pixels on the segmentation result. The average gray value H avg is the average gray value within the neighborhood of the current pixel, which reflects the gray information of the area around the current pixel. And this calculation result is one item in the segmentation unit of the lung and non-lung regions. Combined with the calculation part of "(H i - FGH i )×sgn(H i - FGH i ), a preliminary segmentation result is obtained; The threshold segmentation of this algorithm unit is a simple and efficient segmentation method that can quickly distinguish the lung region from other tissues. The segmentation threshold FGH of the i-th pixel point in the segmentation unit of the lung and non-lung regions i is calculated and set based on the segmentation thresholds of historical patients, which makes the algorithm have strong adaptability and flexibility. In practical applications, it can select appropriate thresholds according to the characteristics of different lung CT images, and thus obtain satisfactory segmentation results; By introducing the neighborhood influence coefficient a, the segmentation unit of the lung and non-lung regions can smoothly segment the edge, reducing the influence of noise on the segmentation result. Moreover, the magnitude of the neighborhood influence coefficient a determines the smoothness of the segmentation edge. When a is large, the segmentation edge is smoother, but some detailed information will be sacrificed. When a is small, the segmentation edge is sharper, but it will be interfered by noise. Therefore, in practical applications, it is necessary to select an appropriate value of a according to the specific situation to obtain a smooth and accurate segmentation result.
[0026] Please refer to Figures 1 to 3 , the system calculates and outputs the gray value H of the i-th pixel point of all pixel points of the patient's lung i and the segmentation result FG of the i-th pixel point i one by one, and performs the operations of dilation and erosion, specifically as follows: ; ; Among them: FG p is the result after dilation, the image after performing the dilation operation on the segmentation result FG of the i-th pixel point i ; FG F is the result after erosion, the image after performing the erosion operation on the segmentation result FG of the i-th pixel point i ; ⊕ represents the dilation operation, represents the erosion operation; YS is the structuring element, and it reflects the pixel area of a small rectangle and a circle; The calculation formula of the morphological processing and segmentation result optimization unit is as follows: FGJ=(FG p -FG F )×β+FG F ; Among them: FGJ is the final segmentation result; β is the weight coefficient of the result FG after dilation p and the result FG after erosion F , and its value range is {0 - 1}, which is used to adjust the influence of dilation and erosion on the final result; The weight coefficient β When the weight coefficient β increases, that is, tends to 1, the contribution of the dilation result to the final segmentation result increases, resulting in the expansion of the lung region; When the weight coefficient β decreases, that is, tends to 0, the contribution of the erosion result to the final segmentation result increases, resulting in the shrinkage of the lung region.
[0027] In this embodiment, this algorithm unit first "FG p-FG F ”The calculation part is to further optimize the segmentation result through morphological processing, namely dilation and erosion. Among them, the dilation operation can expand the target area and fill holes, while the erosion operation can shrink the target area and remove small targets. By calculating “FG p -FG F ”, a difference image reflecting the shape and size changes of the target area can be obtained. And this calculation result is one item in the morphological processing and segmentation result optimization unit. Combined with the subsequent calculation part, the final segmentation result FGJ is obtained. By adjusting the value of β, the influence degree of the dilation and erosion results on the final result can be controlled; “β + FG F ”The calculation part is to add the results of dilation and erosion by weights to obtain the final segmentation result. β is the weight coefficient of the dilation and erosion results, used to adjust the influence of the two on the final result. The result FG F after erosion is the image after the preliminary segmentation result undergoes the erosion operation. It reflects the shape and size of the target area after shrinking processing. And this calculation result is the final segmentation result FGJ in the morphological processing and segmentation result optimization unit, that is, the final segmentation result after morphological processing. By adjusting the value of β, the final segmentation result can be made to more conform to the actual characteristics of the COPD lung region. At the same time, this final result will be used as the input for subsequent processing; In this algorithm unit, through the dilation and erosion operations in morphological processing, small targets can be effectively removed and holes can be filled. By adjusting the weight coefficient β, the influence degree of the dilation and erosion operations on the segmentation result can be controlled. When β is large, the dilation operation dominates and can effectively remove small targets. When β is small, the erosion operation dominates and can fill holes. Therefore, in practical applications, the value of β can be adjusted according to the specific situation of the segmentation result to obtain a more complete and accurate segmentation result. And by adjusting the weight coefficient β, the influence degree of morphological processing on the segmentation result can be flexibly controlled. This flexibility enables the morphological processing and segmentation result optimization unit to be adjusted according to the specific situation of the segmentation result, thereby obtaining a more satisfactory segmentation result. At the same time, morphological processing can further optimize the accuracy of the segmentation result, making the segmentation result more in line with the actual situation; In addition, morphological processing not only optimizes the segmentation result, but also can improve the performance and stability of the algorithm to a certain extent. By removing small targets and filling holes, the redundant information and noise interference in the segmentation result are reduced, thereby improving the running efficiency and stability of the algorithm.
[0028] Please refer to Figures 1 to 3 , the final segmentation result FGJ is used as the input value for subsequent processing, including feature extraction and quality assessment.
[0029] In this embodiment, the morphological processing and segmentation result optimization unit's cyclic influence mechanism on the unit reflecting the brightness characteristics of the lung region enables the entire segmentation system to dynamically adjust according to the quality of the segmentation result. This adaptability and flexibility enable the algorithm to better adapt to the characteristics of different lung CT images, thereby obtaining more accurate and stable segmentation results; The final segmentation result FGJ obtained through morphological processing can be used as the input for subsequent processing. If there are many small targets and hole problems in the segmentation result, the β coefficient can be adjusted to enhance the effect of morphological processing, thereby optimizing the segmentation result. This optimization process will be fed back to the lung and non-lung region segmentation unit, and then the segmentation threshold FGH of the i-th pixel needs to be adjusted. i As well as the neighborhood influence coefficient a to adapt to the new segmentation result. Finally, these adjustments will further affect the gray value calculation in the unit reflecting the brightness characteristics of the lung region, making the entire segmentation system more continuous and consistent; The cyclic influence mechanism of the morphological processing and segmentation result optimization unit on the unit reflecting the brightness characteristics of the lung region not only optimizes the segmentation result, but also can improve the overall performance and stability of the algorithm to a certain extent. By continuously iterating and adjusting parameters, the algorithm can better adapt to different application scenarios and data characteristics, thereby improving the robustness and reliability of the algorithm; In summary, the unit reflecting the brightness characteristics of the lung region, the lung and non-lung region segmentation unit, and the morphological processing and segmentation result optimization unit each have significant beneficial effects. The cyclic influence of the morphological processing and segmentation result optimization unit on the unit reflecting the brightness characteristics of the lung region further enhances the self-adaptability and optimization ability of the entire segmentation system. These beneficial effects act together on the COPD lung region segmentation system based on computer vision, improving the accuracy, stability, and efficiency of the segmentation. In practical applications, appropriate parameters and algorithm steps need to be selected according to specific situations to obtain the best segmentation result.
[0030] Embodiment 2, please refer to Figures 1 to 3 , the segmentation threshold FGH of the i-th pixel i The calculation formula is as follows: ; M is the total number of patients; Among them, according to the segmentation result FG of the i-th pixel of the total number of patients M calculated recently i , collect and calculate, and calculate and output the segmentation threshold FGH of the i-th pixel i As a reference and regularly updated benchmark value; In addition, the segmentation threshold FGH of the i-th pixel needs to be calculated for each pixel in the pixel area of the lung image i for reference.
[0031] In this embodiment, by regularly updating the segmentation threshold FGH of the i-th pixel point i , the system can better adapt to the characteristics of lung images of different patients, including different anatomical structures, pathological changes, and image acquisition conditions. And based on the segmentation data of recent patients for averaging, it can reduce the segmentation error caused by individual differences and image noise, thereby improving the segmentation accuracy; Regularly updating the segmentation threshold FGH of the i-th pixel point i can make the algorithm more robust to noise and artifacts in the image. Because these noises and artifacts will interfere with the segmentation result, and through averaging, it can reduce the fluctuation of the segmentation result caused by abnormal data of a single patient, thereby improving the stability of the algorithm; Regularly updating the segmentation threshold FGH of the i-th pixel point i is actually an adaptive learning process, which enables the algorithm to continuously learn and optimize itself from new data. As the algorithm continues to learn and optimize, its intelligence level will gradually increase, providing a more efficient and accurate solution for future medical image processing; In summary, regularly updating the segmentation threshold FGH of the i-th pixel point for all pixel regions of the lungs i and performing averaging based on the segmentation data of recent patients can significantly improve the segmentation accuracy, enhance the robustness of the algorithm, optimize the segmentation efficiency, and promote the development of algorithm adaptability and intelligence.
[0032] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it is understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. A computer vision-based regional segmentation system for the lung regions of patients with chronic obstructive pulmonary disease, characterized in that, The specific implementation steps are as follows: Step 1: Use the image data acquisition module to obtain the original lung CT images of the patient; Step 2: Use the input module to input the lung CT images into the segmentation processing module; Step 3: The segmentation processing module sequentially calculates and outputs the gray value H of the i-th pixel point i , the segmentation result FG of the i-th pixel point i , and the final segmentation result FGJ; Step 4: Present the final segmentation optimization result of the final segmentation result FGJ on the display module; Among them, the segmentation processing module includes a unit reflecting the brightness characteristics of the lung region, a unit for segmenting the lung and non-lung regions, and a unit for morphological processing and optimizing the segmentation result; When obtaining the original lung CT image of a patient, the gray value H of the i-th pixel point needs to be calculated one by one according to each preset pixel area of the lung i , and based on the gray value H of the i-th pixel point i , judge the segmentation result FG of the i-th pixel point in each pixel area i , and then collect all the segmentation results FG of the i-th pixel point i , and perform dilation and erosion processing to calculate the final segmentation result FGJ.
2. The computer vision-based COPD lung region segmentation system according to claim 1, characterized in that The image data acquisition module includes a CT scanner; The input module includes an input device; The segmentation processing module includes a computer and image processing software; The display module includes a display.
3. The computer vision-based COPD lung region segmentation system according to claim 2, wherein: The calculation formula of the unit reflecting the brightness characteristics of the lung region is as follows: ； Where: H i is the gray value of the i-th pixel point; R i is the red channel value of the i-th pixel, reflecting the red component in the image; G i is the green channel value of the i-th pixel, which reflects the green component in the image and plays a key role in identifying the structure and texture of the lungs in lung CT images; B i is the blue channel value of the i-th pixel, reflecting the green component in the image; is the red channel value R of the i-th pixel i , the green channel value G of the i-th pixel i and the blue channel value B of the i-th pixel i respectively represent different color components and information in the color image. And when processing lung CT images, the information of the above three channels jointly determines the color and brightness of the image.
4. The computer vision-based COPD lung region segmentation system according to claim 3, wherein: The calculation formula of the unit for segmenting the lung and non-lung regions is as follows: FG i =(H i -FGH i )×sgn(H i -FGH i )+a×H avg ; ； Where: FG i is the segmentation result of the i-th pixel, and its value range is {0 - 1}. If it is determined to be 1, it is the lung region; if it is determined to be 0, it is the non-lung region. FGH i is the segmentation threshold for the i-th pixel point; a is the neighborhood influence coefficient, and its value range is {0 - 1}, and it is used to adjust the influence of neighborhood pixels on the segmentation result: When a = 0, it means not considering the neighborhood influence; When a = 1, it means completely relying on neighborhood pixels for segmentation; H avg is the average gray value, reflecting the average degree of gray values within the neighborhood of the current pixel point; N is the total number of pixels, reflecting the total number of all pixels in the neighborhood of the current pixel point; sgn(H i -FGH i ) is specifically the sign function; When H i -FGH i > 0, sgn(H i -FGH i ) = 1; When H i -FGH i ≤ 0, sgn(H i -FGH i ) = -1.
5. The computer vision-based COPD lung region segmentation system according to claim 4, wherein: Not considering the neighborhood influence is as follows: When the image quality is high and the contrast between the lung region and the non-lung region is obvious, the gray value of a single pixel is sufficient for accurate segmentation. At this time, the neighborhood influence is not considered, that is, a = 0 is set; When the processing speed is a key factor and the accuracy requirement for the segmentation result is not high, the neighborhood influence is not considered, that is, a = 0 is set; Completely relying on neighborhood pixels for segmentation is as follows: When the image noise is large and the contrast between the lung region and the non-lung region is low, the gray value of a single pixel is not sufficient for accurate segmentation. At this time, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, a = 1 is set; When processing images with complex textures and structures, the information of neighborhood pixels is used to improve the accuracy of segmentation, that is, a = 1 is set; Among them, according to the set situation, the system automatically adjusts the value of the neighborhood influence coefficient a.
6. The computer vision-based COPD lung region segmentation system according to claim 4, wherein: The gray value H of the i-th pixel point among all pixel points of the patient's lungs in the said system i and the segmentation result FG of the i-th pixel point i are calculated and output one by one, and dilation and erosion operations are performed, specifically as follows: ； ； Where: FG p The result after dilation, which is the segmentation result FG of the i-th pixel i The image after performing the dilation operation; FG F The result after corrosion, the segmentation result FG of the i-th pixel i The image after performing the erosion operation; ⊕ represents the dilation operation, represents the erosion operation; YS is a structural element, and it reflects the pixel area of a small rectangle and a circle.
7. The computer vision-based COPD lung region segmentation system according to claim 6, characterized in that: The calculation formula of the morphological processing and segmentation result optimization unit is as follows: FGJ=(FG p -FG F )×β+FG F ; Where: FGJ is the final segmentation result; β is the result FG after dilation p and the result FG after erosion F is the weight coefficient, and its value range is {0 - 1}, which is used to adjust the influence of dilation and erosion on the final result; The final segmentation result FGJ is used as the input value for subsequent processing, including feature extraction and quality assessment.
8. The computer vision-based COPD lung region segmentation system according to claim 7, characterized in that: Based on the weight coefficient β; When the weight coefficient β increases, that is, approaches 1, the contribution of the dilation result to the final segmentation result increases, resulting in the expansion of the lung region; When the weight coefficient β decreases, that is, approaches 0, the contribution of the erosion result to the final segmentation result increases, resulting in the shrinkage of the lung region.
9. The computer vision-based COPD lung region segmentation system according to claim 7, wherein: The segmentation threshold FGH of the i-th pixel point i has the following calculation formula: ； M is the total number of patients; Among them, according to the segmentation result FG of the i-th pixel point of the total number of patients M output by recent calculations i , collection calculations are performed, and the segmentation threshold FGH of the i-th pixel point is calculated and output i as a reference and regularly updated benchmark value; In addition, the segmentation threshold FGH of the i-th pixel point needs to be calculated for each pixel point in the pixel region of the lung image i for reference.