Tumor CT image intelligent segmentation system based on multimodality and deep learning

Abstract

The invention discloses a tumor CT image intelligent segmentation system based on multimodality and deep learning, and relates to the technical field of image segmentation, the tumor CT image intelligent segmentation system comprises a data collection module, a data marking module, a data processing and adjusting module, a circulation control module and a database module, the data collection module is used for collecting tumor CT images of glioma patients, and the data marking module is used for marking the tumor CT images; the data marking module is used for acquiring x tumor marking CT images which are manually subjected to pixel-level marking, and the data processing and adjusting module is used for acquiring a target segmentation CT image, associated information, regional feature information and feature statistical information, and acquiring a feature weight, a regional feature score and an adjusted segmentation threshold; on the basis of multi-modal data fusion and deep learning, the method can determine the importance of each feature on target region recognition, comprehensively evaluate the matching degree of the region and the target region, and adapt to different image features so as to realize accurate segmentation.

Description

Technical Field

[0001] The present invention relates to the field of image segmentation technology, and specifically to a tumor CT image intelligent segmentation system based on multimodality and deep learning. Background Art

[0002] In the field of tumor diagnosis, accurate CT image segmentation is crucial for tumor detection, localization, and treatment planning. With the development of medical imaging technology, multimodal data provides rich information for tumor analysis. At the same time, deep learning technology has shown great potential in image segmentation and can automatically learn image features to improve segmentation accuracy.

[0003] Currently, intelligent segmentation technology for tumor CT images that combines multimodal data fusion with deep learning has gradually become a research hotspot. However, in practical applications, the intelligent segmentation of most tumor regions is based on manual annotation by multiple experts. There are certain differences in the annotations of different experts. If the largest annotated area is only processed for edge recognition, the accuracy of image segmentation will inevitably be reduced. In addition, when extracting CT image features, existing technologies often fail to fully consider the actual correlation between different features and various tumor areas. As a result, some features that are critical to tumor region identification do not receive the due weight, while some irrelevant or weakly related features may occupy a higher weight, affecting segmentation accuracy.

[0004] Secondly, most existing technologies only rely on a single feature to evaluate whether an image area is a tumor-related area, and thus cannot fully reflect the degree of matching between the area and the tumor area, resulting in inaccurate segmentation results and prone to misjudgment and omission.

[0005] Finally, existing technologies usually use a fixed segmentation threshold to distinguish tumors from non-tumor areas, without dynamically adjusting it based on the feature distribution of different areas in the image. This makes the segmentation threshold unable to adapt to the feature differences of different CT images, and the segmentation effect is poor when faced with complex and changeable tumor images. Summary of the Invention

[0006] The purpose of the present invention is to provide an intelligent tumor CT image segmentation system based on multimodality and deep learning, which solves the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution, including a data collection module, a data marking module, a data processing and adjustment module, a cycle control module and a database module; The data collection module is used to collect tumor CT images of glioma patients that meet the MNI standard and are converted to the MNI standard space, and transmit them to the data marking module; The data labeling module is used to obtain x tumor-annotated CT images that are manually annotated at the pixel level on the multimodal three-dimensional images; The pixel-level annotation includes background, edema area, necrosis area and enhancement area; Extracting a target segmented CT image containing target and non-target image areas according to diagnostic setting features preset by the system, and acquiring feature statistical information and regional feature information in the target segmented CT image using the data collection module; Segmenting the CT image according to the target to obtain related information; The data processing and adjustment module is used to obtain the target segmented CT image, the correlation information, the regional feature information and the feature statistical information; Obtaining feature weights according to the association information and the feature statistical information; Obtaining a regional feature score according to the feature weight and the regional feature information; Obtaining an adjusted segmentation threshold according to all the regional feature scores in the target segmented CT image and the segmentation adjustment basis threshold and adjustment coefficient obtained from the database module; The loop control module is configured to receive the adjusted segmentation threshold and the regional feature score, and determine and intelligently execute a segmentation result based on a comparison result of the adjusted segmentation threshold and the regional feature score.

[0008] Optionally, the equipment used by the data collection module includes a CT scanner and clinical examination instruments; The equipment used by the data marking module includes image annotation software; The equipment used by the data processing and adjustment module and the cycle control module includes a processor; The equipment used by the database module includes a storage server; The database module stores and transmits the data collection module, the data marking module, the data processing and adjustment module and the cycle control module through a data interface; The purpose of database module design is not only to realize the storage and management of related data, but also to form a complete data model for the originally loose and independent data.

[0009] Optionally, the diagnostic setting features include inflammation diagnostic features and growth metabolism diagnostic features; The inflammation diagnostic feature uses the inflammatory response of the tissue surrounding the tumor reflected by the edema area as the target diagnostic feature; The growth and metabolism diagnostic features use the growth and metabolism of the tumor reflected by the necrotic area as the target diagnostic features.

[0010] Optionally, the feature statistical information includes the number of occurrences of a single feature and the total number of occurrences of features; Obtaining a feature proportion of the single feature in the target segmented CT image according to a proportional relationship between the number of occurrences of the single feature and the number of occurrences of the total features; In tumor CT image segmentation with multimodal data fusion, different features have different effects on distinguishing target and non-target image areas of the tumor. The feature proportion obtained based on the feature statistical information is multiplied by the correlation information to obtain the feature weight of each feature that can accurately quantify the importance of each feature to target area identification.

[0011] Optionally, the regional feature information includes an average value of a single feature in the region and the feature weight of each feature; The details are as follows: After matching the association information and the feature statistical information of each target and non-target image area for the same feature in the same area, obtain the feature weight of each feature in each area; Segment the CT image according to the target, and obtain intra-regional feature quantities and intra-regional multi-feature information of each target and non-target image region; The multi-feature information in the region includes grayscale features, texture features and shape features; Based on the multi-feature information in the region, the value of the feature quantity in the region is 3; After averaging the feature quantity in the region and the multi-feature information in the region to match each target and non-target image region, an average value of the single feature in each region is obtained; In tumor CT images, whether a region is a tumor-related region requires a comprehensive multi-feature judgment. The average value of the single feature of the region, the feature quantity within the region, and the current regional feature weight obtained based on the feature weight and the regional feature information are finally obtained to obtain the regional feature score under the multi-feature evaluation of each region to comprehensively evaluate the degree of matching between the region and the target region.

[0012] Optionally, based on the regional feature scores of all regions obtained for each target and non-target image region, a maximum regional feature score and an average regional feature score are obtained; Obtaining a segmentation adjustment amount according to the maximum regional feature score, the average regional feature score, and the adjustment coefficient; Adding the segmentation adjustment basis threshold to the segmentation adjustment amount obtained according to the regional feature scores of all regions and the adjustment coefficient, and finally obtaining the adjusted segmentation threshold; The adjustment coefficient ranges from 0 to 1. The segmentation adjustment threshold is set to 0.5 during the initial segmentation, and after the initial segmentation, it is set to the last adjusted segmentation threshold.

[0013] Optionally, the data processing and adjustment module includes a feature processing unit and a threshold adjustment unit; The cycle control module includes a segmentation determination unit, an image segmentation unit and a cycle update unit; Wherein, the feature processing unit is used to obtain the feature weight and the regional feature score; The threshold adjustment unit is used to obtain the adjusted segmentation threshold; The segmentation determination unit is used to obtain the comparison result and the segmentation result; The image segmentation unit and the loop updating unit are used to execute the segmentation result.

[0014] Optionally, the comparison results specifically include the following: S1: the regional feature score is greater than the adjusted segmentation threshold; S2: the regional feature score is less than or greater than the adjusted segmentation threshold; The segmentation results based on S1 and S2 are as follows: Based on S1, determine and segment the current area into the target image area; Based on S2, determine and segment the current area as a non-target image area; The loop control module stops looping and segmenting when the segmentation results of two adjacent times are the same.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention fully considers the feature proportion of features in feature statistical information and the correlation information with the target area. In practical applications, for features that are common in tumor CT images and closely related to the tumor area, by counting the number of occurrences of single features and the total number of occurrences of features, and combining the correlation information determined by the machine learning algorithm, the feature weight of each feature in each area can be scientifically allocated, and the features that are important for tumor area identification can be more accurately highlighted, avoiding interference from irrelevant or weakly correlated features, thereby improving the accuracy of image segmentation.

[0016] When processing tumor CT images, the regional feature score evaluates each region from multiple feature dimensions, taking into account grayscale features, texture features, and shape features. This comprehensive evaluation method can more accurately reflect the degree of match between the region and the tumor region, reduce misjudgments and missed judgments, and make the segmentation results more consistent with the actual tumor area distribution.

[0017] Finally, this system dynamically adjusts the segmentation threshold based on the maximum regional feature score and the average regional feature score of all regions and the adjustment coefficient. Therefore, when faced with tumor CT images of different patients, the system can automatically adjust the segmentation threshold according to the distribution of regional scores in the image. This dynamic adjustment mechanism enables the system to adapt to various complex and changeable tumor CT images and improve the stability and reliability of segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is the overall module flow chart of the intelligent segmentation system for tumor CT images; Figure 2 Schematic diagram of the adjustment acquisition of the data processing and adjustment module of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] Regarding this tumor CT image intelligent segmentation system, it is different from the existing tumor CT image intelligent segmentation system; The existing intelligent segmentation system of tumor CT images has problems such as unreasonable feature extraction and weight distribution, single and inaccurate regional assessment, and lack of dynamic adjustment of segmentation threshold setting. However, this algorithm unit can perform scientific feature weight calculation, comprehensive and accurate regional assessment, and dynamic and adaptive segmentation threshold adjustment.

[0021] For example 1, please refer to Figures 1 to 2 ,This implementation provides a tumor CT image intelligent segmentation system based on multimodality and deep learning, including a data collection module, a data labeling module, a data processing and adjustment module, a loop control module, and a database module; A data collection module is used to collect tumor CT images of glioma patients and transmit them to a data labeling module; The data labeling module is used to obtain x tumor annotated CT images with manual pixel-level annotation; Among them, pixel-level annotation includes background, edema area, necrosis area and enhancement area; According to the diagnostic setting features pre-set by the system, the target segmented CT image containing the target and non-target image areas is extracted, and the feature statistical information and regional feature information in the target segmented CT image are obtained by using the data collection module; Diagnostic setting features include inflammatory diagnostic features and growth and metabolism diagnostic features; The inflammatory diagnostic feature uses the inflammatory response of the tissue surrounding the tumor reflected by the edema area as the target feature for diagnosis; Growth and metabolism diagnostic features take the growth and metabolism of the tumor reflected by the necrotic area as the target diagnostic features; Segment the CT image according to the target and obtain related information; Data processing and adjustment module, used to obtain target segmentation CT images, correlation information, regional feature information and feature statistical information; Obtain feature weights based on correlation information and feature statistics; Obtain regional feature scores based on feature weights and regional feature information; Obtaining an adjusted segmentation threshold according to all regional feature scores in the target segmented CT image and the segmentation adjustment basis threshold and adjustment coefficient obtained from the database module; a loop control module, configured to receive the adjusted segmentation threshold and the regional feature score, and determine and intelligently execute the segmentation result based on a comparison result of the adjusted segmentation threshold and the regional feature score; The data processing and adjustment module includes a feature processing unit and a threshold adjustment unit; The loop control module includes a segmentation determination unit, an image segmentation unit and a loop update unit; Among them, the feature processing unit is used to obtain feature weights and regional feature scores; The threshold adjustment unit is used to obtain the adjusted segmentation threshold; The segmentation judgment unit is used to obtain comparison results and segmentation results; The image segmentation unit and the loop updating unit are used to execute the segmentation results.

[0022] The equipment used in the data collection module includes CT scanners and clinical examination instruments; The equipment used in the data labeling module includes image annotation software; The equipment used by the data processing and adjustment module and the cycle control module includes a processor; The equipment used by the database module includes a storage server; The database module stores and transmits the data to the data collection module, the data marking module, the data processing and adjustment module and the cycle control module through the data interface.

[0023] In this embodiment, the system requires x, or more than 15, experts with more than three years of clinical experience to manually label multimodal 3D images, especially pixel-level annotation. Manual labeling leverages the experts' expertise and clinical experience to accurately identify the category of each pixel in the image, providing reliable basic data for subsequent intelligent processing. Afterwards, the overlapping areas of all manually annotated regions are extracted. Since there are certain differences in the annotations of different experts, the overlapping areas represent relatively consistent judgments. This serves as the basic data for intelligent processing and can reduce errors caused by annotation differences. Next, based on the target segmentation of the extracted overlapping regions, feature weights are sequentially obtained that accurately reflect the importance of each feature in distinguishing the target region, regional feature scores that enable the regional score to more realistically reflect the degree of match between the region and the target region, and regional feature scores that are more suitable for image segmentation based on overlapping region features. In summary, this method of first manually labeling and then intelligently processing overlapping areas fully utilizes the advantages of manual professional judgment and intelligent algorithm data processing. Manual labeling provides a professional and accurate data basis, and intelligent processing uses algorithms to efficiently analyze and calculate large amounts of data to achieve automation and intelligence of image segmentation. Moreover, for the development of intelligent diagnosis products for gliomas, accurate image segmentation is a key link. Through the fusion of manual and intelligent processing, the accuracy of image segmentation can be improved, providing high-quality data support for the subsequent development of intelligent diagnosis products.

[0024] See also Figures 1 to 2 ,Feature statistics include the number of occurrences of a single feature and the number of occurrences of the total features; According to the ratio between the number of occurrences of a single feature and the number of occurrences of the total features, the feature proportion of the single feature in the target segmented CT image is obtained; Multiply the feature proportion obtained from the feature statistical information by the associated information to obtain the feature weight of each feature; In this embodiment, the calculation formula of the feature weight is as follows: ; in: Q i is the feature weight; T i is the number of occurrences of the single feature of the i-th feature in the total features, T i It reflects the frequency of occurrence of the feature in the target segmentation CT image. Features with more occurrences occupy a more important position in the overall total features and may also have a closer connection with the target area. ZT is the total number of feature occurrences; G i To correlate information, the logistic regression model can be used to directly and intelligently obtain it; The calculation result is the feature proportion. The higher the value of the feature proportion, the more common the feature is in the overall total features.

[0025] It can effectively integrate feature information from different modal data. Specifically, features extracted from multi-dimensional information of clinical data, imaging data and pathological data can be used to Calculate its feature weight Q i , so that on the basis of multimodal data fusion, each feature can find a suitable position in the overall segmentation algorithm, laying the foundation for subsequent accurate segmentation based on multimodal data.

[0026] See also Figures 1 to 2 , the regional feature information includes the average value of the regional single feature and the feature weight of each feature; The details are as follows: After matching the association information and feature statistical information of each target and non-target image area with the same feature in the same area, obtain the feature weight of each feature in each area; Segment the CT image according to the target, and obtain the intra-regional feature quantity and intra-regional multi-feature information of each target and non-target image region; The multi-feature information in the region includes grayscale features, texture features and shape features; Based on the multi-feature information within the region, the value of the feature quantity within the region is 3; After averaging the feature quantity and multi-feature information within the region and matching each target and non-target image region, the average value of the single feature of each region is obtained; Finally, obtaining the regional feature score of each region based on the feature weight and the regional feature information obtained from the regional single feature average, the feature quantity within the region, and the current regional feature weight; In this embodiment, the calculation formula of the regional feature score is as follows: ; in: QY i is the regional feature score, that is, the sum of the contribution values ​​of all features in the jth region score; n is the characteristic quantity in the region, Q ij is the feature weight of the single feature of the jth region, H ij is the average value of a single feature in j regions.

[0027] In tumor CT images, whether an area belongs to a tumor-related area cannot be judged based on a single feature alone, but requires comprehensive consideration of multiple features. In this embodiment, not only the grayscale value feature is considered, but also the texture feature and shape feature are combined, and the feature weight Q of each feature is used to determine whether an area belongs to a tumor-related area. iThe weighted summation of multiple features comprehensively evaluates the degree of match between the area and the target area, greatly improving the accuracy of judgment compared to the evaluation method that relies solely on a single feature.

[0028] See also Figures 1 to 2 , based on the regional feature scores of all regions obtained for each target and non-target image region, obtain the maximum regional feature score and the average regional feature score; Obtain the segmentation adjustment amount according to the maximum score of the regional characteristics, the average score of the regional characteristics and the adjustment coefficient; Add the segmentation adjustment basis threshold to the segmentation adjustment amount obtained according to the regional feature scores of all regions and the adjustment coefficient, and finally obtain the adjusted segmentation threshold; Among them, the value range of the adjustment coefficient is 0-1; The segmentation adjustment threshold is set to 0.5 at the initial segmentation, and after the initial segmentation, it is set to the last adjusted segmentation threshold; In this embodiment, the calculation formula of the segmentation threshold after adjustment is as follows: ; in: FT new is the adjusted segmentation threshold, FT old Adjust the threshold value for segmentation, QY avg is the average score of regional characteristics, QY max is the maximum score of regional characteristics, k is the adjustment coefficient; The result of calculation is the split adjustment amount; By adjusting the segmentation threshold FT new The system can obtain the average score QY based on the regional features of the current image. avg and the maximum score of regional features QY max The segmentation threshold is automatically adjusted based on the scores of the composed regions; For images with obvious tumor features and large regional score differences, the adjusted threshold can more accurately divide tumor and non-tumor areas; For images with more complex features and smaller regional score differences, the appropriate threshold can also be adaptively adjusted to improve the accuracy and adaptability of segmentation.

[0029] It is worth noting that in the intelligent segmentation of tumor CT images, the characteristic distribution of the data and the stability of the segmentation results are important considerations. A smaller adjustment coefficient k, that is, k close to 0, will make the adjustment of the segmentation threshold smaller. In this way, under the premise of ensuring the relative stability of the segmentation results, the average score QY of the regional features can be used to calculate the segmentation threshold. avg and the maximum score of regional features QY maxFine-tune the scores of the composed regions to avoid destroying the existing segmentation effect due to excessive adjustments; A larger adjustment coefficient k, that is, k close to 1, will increase the adjustment amplitude, but too large a value will cause excessive fluctuations in the segmentation results, which is not conducive to obtaining stable and accurate segmentation results. Therefore, the value range of k between 0 and 1 can find a balance between the adaptability of data features and segmentation stability.

[0030] For example 2, please refer to Figures 1 to 2 The comparison results are as follows: S1: The regional feature score is greater than the adjusted segmentation threshold; S2: The regional feature score is less than or greater than the adjusted segmentation threshold; The segmentation results based on S1 and S2 are as follows: Based on S1, determine and segment the current area into the target image area; Based on S2, determine and segment the current area as a non-target image area; The loop control module stops looping and segmenting when the segmentation results of two adjacent segments are the same.

[0031] In this embodiment, tumors are highly heterogeneous, and tumor characteristics may vary between different patients or even in different parts of the same patient. The cyclic influence of the cyclic control module enables the system to continuously adapt to this heterogeneity in multiple iterations. As the segmentation threshold is adjusted, the feature weights also change accordingly, enabling the algorithm to better capture the various complex characteristics of the tumor, improve the segmentation capabilities of different tumor types and morphologies, and enhance the generalization and adaptability of the algorithm.

[0032] This cyclic influence mechanism makes the entire segmentation system adaptive and intelligent. The system can dynamically adjust the feature weights according to each segmentation result without manual intervention. When faced with a large number of different tumor CT images, the system can automatically learn the changing patterns of image features, continuously optimize its own segmentation strategy, and improve segmentation efficiency and accuracy.

[0033] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. Intelligent tumor CT image segmentation system based on multimodality and deep learning, characterized by: It includes data collection module, data marking module, data processing and adjustment module, cycle control module and database module; The data collection module is used to collect tumor CT images of glioma patients and transmit them to the data marking module; The data labeling module is used to obtain x tumor-labeled CT images that are manually labeled at the pixel level; The pixel-level annotation includes background, edema area, necrosis area and enhancement area; Extracting a target segmented CT image containing target and non-target image areas according to diagnostic setting features preset by the system, and acquiring feature statistical information and regional feature information in the target segmented CT image using the data collection module; Segmenting the CT image according to the target to obtain related information; The data processing and adjustment module is used to obtain the target segmented CT image, the correlation information, the regional feature information and the feature statistical information; Obtaining feature weights according to the association information and the feature statistical information; Obtaining a regional feature score according to the feature weight and the regional feature information; Obtaining an adjusted segmentation threshold according to all the regional feature scores in the target segmented CT image and the segmentation adjustment basis threshold and adjustment coefficient obtained from the database module; The loop control module is configured to receive the adjusted segmentation threshold and the regional feature score, and determine and intelligently execute a segmentation result based on a comparison result of the adjusted segmentation threshold and the regional feature score.

2. The intelligent tumor CT image segmentation system based on multimodality and deep learning according to claim 1 is characterized in that: The equipment used in the data collection module includes CT scanners and clinical examination instruments; The equipment used by the data marking module includes image annotation software; The equipment used by the data processing and adjustment module and the cycle control module includes a processor; The equipment used by the database module includes a storage server; The database module stores and transmits the data collection module, the data marking module, the data processing and adjustment module and the cycle control module through a data interface.

3. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 1, characterized in that: The diagnostic setting features include inflammation diagnostic features and growth metabolism diagnostic features; The inflammation diagnostic feature uses the edema area as the target feature for diagnosis; The growth metabolism diagnostic feature uses the necrotic area as the target feature for diagnosis.

4. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 3 is characterized by: The feature statistics include the number of occurrences of a single feature and the total number of occurrences of features; Obtaining a feature proportion of a single feature based on a proportional relationship between the number of occurrences of the single feature and the number of occurrences of the total features; The feature proportion obtained according to the feature statistical information is multiplied by the associated information to obtain the feature weight of each feature.

5. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 4 is characterized by: The regional feature information includes the average value of a single feature of the region and the feature weight of each feature; The details are as follows: After matching the association information and the feature statistical information of each target and non-target image area for the same feature in the same area, obtain the feature weight of each feature in each area; Segment the CT image according to the target, and obtain intra-regional feature quantities and intra-regional multi-feature information of each target and non-target image region; The multi-feature information in the region includes grayscale features, texture features and shape features; Based on the multi-feature information in the region, the value of the feature quantity in the region is 3; After averaging the feature quantity and the regional feature matching of each target and non-target image area, an average value of the regional single feature of each area is obtained; The regional feature score of each region is finally obtained according to the feature weight and the regional single feature average value obtained from the regional feature information, the feature quantity within the region and the current regional feature weight.

6. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 5, characterized in that: Obtaining a maximum regional feature score and an average regional feature score based on the regional feature scores of all regions obtained for each target and non-target image region; Obtaining a segmentation adjustment amount according to the maximum regional feature score, the average regional feature score, and the adjustment coefficient; Adding the segmentation adjustment basis threshold to the segmentation adjustment amount obtained according to the regional feature scores of all regions and the adjustment coefficient, and finally obtaining the adjusted segmentation threshold; The adjustment coefficient ranges from 0 to 1. The segmentation adjustment basis threshold is set to 0.5 during the initial segmentation, and after the initial segmentation, it is set to the last adjusted segmentation threshold.

7. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 6, characterized in that: The data processing and adjustment module includes a feature processing unit and a threshold adjustment unit; The cycle control module includes a segmentation determination unit, an image segmentation unit and a cycle update unit; Wherein, the feature processing unit is used to obtain the feature weight and the regional feature score; The threshold adjustment unit is used to obtain the adjusted segmentation threshold; The segmentation determination unit is used to obtain the comparison result and the segmentation result; The image segmentation unit and the loop updating unit are used to execute the segmentation result.

8. The multimodal and deep learning-based intelligent tumor CT image segmentation system according to claim 7, characterized in that: The comparison results specifically include the following: S1: the regional feature score is greater than the adjusted segmentation threshold; S2: the regional feature score is less than or greater than the adjusted segmentation threshold; The segmentation results based on S1 and S2 are as follows: Based on S1, determine and segment the current area into the target image area; Based on S2, determine and segment the current area as a non-target image area; The loop control module stops looping and segmenting when the segmentation results of two adjacent times are the same.

Images