CT image quantitative analysis system based on machine learning

Abstract

The invention discloses a CT image quantitative analysis system based on machine learning, and relates to the technical field of image processing. The information acquisition module is used for acquiring CT image data and analysis requirements; and the information processing module is used for performing specification unified processing on the CT image data through a specification processing method based on the CT image data to obtain to-be-analyzed image data. Through cooperation of the preset verification number and the subsequent cross verification module, cross verification optimization can be performed on multiple pieces of sub-information obtained from the to-be-analyzed image data, so that the accuracy of reflecting the name and number information of the feature region of the to-be-analyzed image data is improved, and the precision of subsequent model processing is improved; according to the method, the efficiency of feature recognition analysis can be improved by analyzing a single feature region through a single feature recognition model, and meanwhile, a plurality of different recognition analysis requirements can be processed, and multi-task processing is embodied, so that the efficiency of information processing is improved, and the increasing clinical precision diagnosis requirements can be met.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of image processing, in particular to a CT image quantitative analysis system based on machine learning. BACKGROUND

[0002] The CT image quantitative analysis system based on machine learning is an intelligent platform combining medical image processing, artificial intelligence and machine learning technology, aiming to improve the diagnostic efficiency and accuracy of medical images. It extracts rich feature information from CT images through automatic and quantitative methods, supports early diagnosis of diseases, treatment evaluation and disease monitoring.

[0003] The patent with the publication number CN112037218A discloses a new coronavirus infection quantitative analysis method based on CT images, which relates to the technical field of image processing and comprises the following steps: S1: establishing a machine learning model; S2: inputting CT images for processing; S3: generating 3D lesion data and calculating a new coronavirus infection quantitative factor; S4: determining whether the quantitative factor score is higher than the critical threshold; if yes, the patient is a new crown patient and S5 is executed; otherwise, the patient does not have new coronavirus infection; S5: repeatedly executing S2 to S4, and after accumulating the quantitative factors of the patient's multiple CT images, the patient's disease trend is obtained. The new coronavirus infection quantitative analysis method based on CT images is simple and convenient, generates 3D lesion data from CT images, scores by calculating and quantifying the lesions, and doctors can directly observe the lesion position, size and changes, etc. information, which can effectively determine whether the patient has new coronavirus infection and further determine the disease trend of the new crown patient.

[0004] In existing CT image analysis, although most systems have automatic recognition and quantitative functions, they mostly rely on a single model or rule, and when facing complex and fuzzy image regions, misjudgment and missed detection may occur. This method relying on a single information source lacks multi-angle and multi-level verification mechanisms, making it difficult to ensure high precision and high robustness. At the same time, when a single model analyzes and processes feature information of multiple feature regions, the efficiency is low, and in the absence of cross-validation and multi-task processing environment, it is difficult to meet the growing demand for precise diagnosis in clinical practice. Therefore, the present application is proposed. SUMMARY

[0005] The purpose of the present application is to provide a CT image quantitative analysis system based on machine learning to solve the problems raised in the background.

[0006] To achieve the above purpose, the present application provides the following technical solution: a CT image quantitative analysis system based on machine learning, comprising:

[0007] An information acquisition module acquires CT image data and analysis requirements;

[0008] The information processing module: based on the CT image data, the CT image data is uniformly processed by the specification processing method to obtain the to-be-analyzed image data;

[0009] It is characterized by comprising:

[0010] The region acquisition module: preset verification quantity, based on the verification quantity, the feature region name and the quantity information of the to-be-analyzed image data are obtained to obtain the result information, the result information includes several sub-information, and the number of sub-information is consistent with the verification quantity;

[0011] The cross-validation module: based on the result information, the cross-validation method is used to judge whether the sub-information of the to-be-analyzed image data exists difference to obtain the verification result, and based on the verification result, the sub-information with difference is judged by the auxiliary method to generate feedback information and optimization information;

[0012] The multi-task processing module: preset basic feature region, a storage library for storing different feature regions and their feature information is established by an information acquisition method, a feature recognition model for analyzing different feature regions is established based on the storage library, an association relationship between the basic feature region and the feature recognition model is established, and a relationship directory is established for storing the basic feature region, the feature recognition model and the association relationship;

[0013] The demand matching module: based on the analysis demand, the corresponding feature recognition model of the analysis demand is found in the relationship directory by the matching method to obtain the target recognition model, the image data related to the analysis demand in the to-be-analyzed image data is labeled based on the optimization information to obtain the target image data, and the target image data is imported into the target recognition model to obtain the analysis information.

[0014] Further, the cross-validation method comprises: splitting the result information to obtain several sub-information, extracting the feature region information in the sub-information to obtain sub-feature information, extracting the feature region quantity information in the sub-information to obtain sub-quantity information, comparing whether the sub-feature information in the several sub-information is consistent to obtain a first comparison result, comparing whether the sub-quantity information in the several sub-information is consistent to obtain a second comparison result, obtaining a comparison result based on the first comparison result and the second comparison result, when the first comparison result and the second comparison result feedback that there is inconsistent information in the several sub-information, the comparison result feedback is that there is difference, extracting the inconsistent information items in the several sub-information to obtain the to-be-verified information, and integrating the comparison result and the to-be-verified information to obtain the verification result.

[0015] Furthermore, the auxiliary method includes: splitting the verification result to obtain the information to be verified, presetting the selection object, the selection object includes quantity selection, specified selection and auxiliary selection, determining the inconsistent information items in the information to be verified based on the selection object to obtain the replacement information item, and replacing the part of the sub-information corresponding to the replacement information item with the replacement information item to obtain optimized information.

[0016] Furthermore, the process of establishing a feature recognition model for analyzing features of different basic parts based on the repository is as follows: Data acquisition and processing: determining the recognition object of the feature recognition model to obtain the target object, traversing the repository based on the target object to select the sub-repository corresponding to the target object to obtain the target repository, and extracting reference information in the target repository to obtain training data;

[0017] Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for model training to obtain the initial model;

[0018] Model adjustment and output: Preset verification images and verification information, import the verification images into the initial model to obtain test information, compare the test information with the verification information to obtain the difference results, and adjust and optimize the initial model based on the difference results to obtain the feature recognition model of the target object;

[0019] Integrated output: Integrate the feature recognition models of all target objects to complete the establishment of feature recognition models for analysis of different feature areas.

[0020] Furthermore, the information acquisition method includes: establishing a sub-repository with the basic feature area as the name, establishing a repository for storing the sub-repository, obtaining CT image information and feature information of the basic feature area to obtain reference information, and storing the reference information in the sub-repository.

[0021] Furthermore, the matching method includes: splitting and analyzing the requirements to obtain the target feature area and information requirements, traversing the relationship directory based on the target feature area to obtain the traversal result, selecting the feature recognition model based on the traversal result and the association relationship to obtain the selected recognition model, and adjusting the output mode of the selected recognition model based on the information requirement to obtain the target recognition model.

[0022] Further, the process of obtaining the result information based on the verification quantity and the feature region name and quantity information of the to-be-analyzed image data is as follows: determining a target path for obtaining the feature region name and quantity information of the to-be-analyzed image data, obtaining an identification accuracy of the target path, ranking the target paths in descending order of the identification accuracy to obtain a ranking result, selecting a target path consistent with the verification quantity from the ranking result in descending order of the identification accuracy to obtain a plurality of selected paths, obtaining sub-information reflecting the feature region name and quantity information of the to-be-analyzed image data based on the plurality of selected paths and the to-be-analyzed image data, and integrating the plurality of sub-information to obtain the result information.

[0023] Further, the specification processing method comprises: presetting a target format, pre-processing the CT image data to obtain pre-processed CT image data, determining whether the CT image data is consistent with the target format to obtain a determination result, when the determination result feedback is that the CT image data is inconsistent with the target format, scaling the CT image data horizontally and vertically to make it consistent with the target format to obtain to-be-analyzed image data and target scaling information, and obtaining actual information based on the target scaling information and analysis information.

[0024] Compared with the prior art, the beneficial effects of the present application are:

[0025] The CT image quantitative analysis system based on machine learning can cross-verify a plurality of sub-information obtained from the to-be-analyzed image data through the preset verification quantity and the subsequent cross-validation module, and optimize the sub-information, so as to improve the accuracy of the feature region name and quantity information reflecting the to-be-analyzed image data, and further improve the accuracy of subsequent model processing. Through a single feature recognition model, only a single feature region is analyzed, which can improve the efficiency of feature recognition analysis, and can process multiple different recognition analysis requirements, embody multi-task processing, and improve the efficiency of information processing, so as to meet the increasing demand for precise diagnosis in clinical practice.

[0026] At the same time, the cross-validation method can be used to determine whether the sub-information of the to-be-analyzed image data has differences. When the verification result feedback is that the plurality of sub-information has differences, the sub-information is judged and optimized by the auxiliary method to generate feedback information and optimization information. The optimization information is the optimized sub-information, and the feedback information is the sub-information with differences. By feeding back the feedback information to the manager, the manager can optimize and adjust the target path for generating the sub-information, so as to improve the accuracy of data processing.

[0027] Meanwhile, the CT image data of different specifications are uniformly processed through the specification processing method, so as to improve the analysis efficiency of the analysis system, the standardized image data can improve the stability and generalization ability of the model, avoid errors caused by data differences, and based on the information requirement, the output mode of the recognition model is selected to obtain the target recognition model, that is, the information format output by the target recognition model is adjusted according to the information format requirement, so that the information format output meets the requirement of the staff. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 A brief flow structure schematic diagram of the present application is shown in the figure.

[0029] Figure 2 A main flow structure schematic diagram of the present application is shown in the figure.

[0030] Figure 3 A cross-validation method structure schematic diagram of the present application is shown in the figure.

[0031] Figure 4 A labeling process structure schematic diagram of the present application is shown in the figure.

[0032] Figure 5 A feature recognition model training structure schematic diagram of the present application is shown in the figure.

[0033] Figure 6 A specification processing method structure schematic diagram of the present application is shown in the figure. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0035] CT image quantitative analysis is a process of using computer technology to quantitatively measure and extract features from CT (Computed Tomography) images. It converts the information in the image into numerical indicators, helping doctors to more objectively and accurately evaluate the state of tissues, organs and lesions, thereby supporting diagnosis and treatment decisions.

[0036] As shown in Figures 1-6 The present application provides a technical solution: a CT image quantitative analysis system based on machine learning, which comprises:

[0037] An information acquisition module: acquiring CT image data and analysis requirements;

[0038] It should be noted that the process of obtaining CT image data includes connecting the diagnosis and treatment system, screening target images, ensuring quality, and standardizing formats and data collection. The analysis requirements are obtained through clinical communication or system input, and the target site and indicators are determined to ensure accurate expression of requirements.

[0039] The information processing module: based on the CT image data, the CT image data is uniformly processed by the specification processing method to obtain the analyzed image data;

[0040] It should be noted that the CT image data of different specifications is uniformly processed by the set specification processing method, so as to improve the analysis efficiency of the analysis system. The standardized image data can improve the stability and generalization ability of the model, and avoid errors caused by data differences.

[0041] It is characterized by comprising:

[0042] The region acquisition module: preset verification quantity, based on the verification quantity, the feature region name and quantity information of the analyzed image data are obtained to obtain the result information, the result information includes several sub-information, and the number of sub-information is consistent with the verification quantity;

[0043] It should be noted that the verification quantity is the number of the whole sub-information, and a single sub-information is obtained by a single way cooperating with the analyzed image data. Through the preset verification quantity cooperating with the subsequent cross-validation module, the sub-information obtained from the analyzed image data can be cross-validated and optimized, so as to improve the accuracy of the feature region name and quantity information of the analyzed image data, and further improve the accuracy of the subsequent model processing.

[0044] The cross-validation module: based on the result information, the cross-validation method is used to judge whether the sub-information of the analyzed image data has difference to obtain the verification result, and based on the verification result, the auxiliary method is used to judge the sub-information with difference to generate feedback information and optimization information;

[0045] It should be noted that the cross-validation method can be used to judge whether the sub-information of the analyzed image data has difference. When the verification result feedback is that several sub-information has difference, the auxiliary method is used to judge and optimize the sub-information to generate feedback information and optimization information. The optimization information is the optimized sub-information, and the feedback information is the sub-information with difference. By feeding back the feedback information to the manager, the manager can correspondingly optimize and adjust the way of generating sub-information, so as to improve the accuracy of data processing.

[0046] The multi-task processing module: a storage library for storing different feature regions and their feature information is established by an information acquisition method, a feature recognition model for analyzing different feature regions is established based on the storage library, an association relationship between the basic feature region and the feature recognition model is established, and a relationship directory is established for storing the basic feature region, the feature recognition model and the association relationship;

[0047] It should be noted that the storage library is established by the set information acquisition method, the storage library includes several sub-storage libraries, the information of a single feature region is stored in the sub-storage library, so as to establish a feature recognition model of a single feature region according to a single sub-storage library, and only a single feature region is analyzed through a single feature recognition model, which can improve the efficiency of feature recognition analysis, and can process multiple different recognition analysis requirements, embody multi-task processing, improve the efficiency of information processing, and lock the corresponding feature recognition model according to the basic feature region through the establishment of the association relationship, and through the establishment of the relationship directory, the searching work in the relationship directory can be facilitated.

[0048] The demand matching module: based on the analysis demand, a target recognition model is found in the relationship directory by a matching method, target image data is obtained by marking the image data related to the analysis demand in the analysis image data based on the optimization information, and the analysis information is obtained by importing the target image data into the target recognition model.

[0049] It should be noted that the target recognition model is found in the relationship directory according to the analysis demand and the matching method, the analysis information is obtained by importing the target recognition model after marking the analysis image data in the optimization information and the analysis demand, and the analysis information can be used to assist medical staff in making judgments.

[0050] In the specific implementation process, for example, Figure 4As shown, first, the CT image data to be processed is loaded into the analysis environment, and the feature information of the target region is obtained or predefined. The feature information of the target region is defined by analyzing the requirements, such as intensity value range, texture feature, boundary characteristic, shape parameter, etc. These feature information supports subsequent screening and matching. A region segmentation technique (such as threshold segmentation, region growing, clustering, multi-threshold method, etc.) is used to extract multiple candidate regions from the original image. The purpose of this step is to distinguish the potential target region from the background and provide a basis for feature comparison. For each candidate region, its feature parameters are extracted and then compared with the known target feature information. According to the feature similarity, regions that meet certain conditions are screened out. These regions are identified as potential target regions, and the regions that meet the conditions are marked on the original image. The marking methods can include adding a boundary contour, changing the region color, and adding a marker point or label in the center of the region. Finally, the target region is clearly marked in the image, providing a basis for subsequent quantitative analysis, report generation, or auxiliary diagnosis.

[0051] As shown in Figure 3 The cross-validation method includes: splitting the result information to obtain a plurality of sub-information, extracting the feature region information in the sub-information to obtain sub-feature information, extracting the feature region quantity information in the sub-information to obtain sub-quantity information, comparing whether the sub-feature information in the plurality of sub-information is consistent to obtain a first comparison result, comparing whether the sub-quantity information in the plurality of sub-information is consistent to obtain a second comparison result, obtaining a comparison result based on the first comparison result and the second comparison result, when the first comparison result and the second comparison result feedback that there is inconsistent information in the plurality of sub-information, the comparison result feedback is that there is a difference, extracting the inconsistent information items in the plurality of sub-information to obtain to-be-verified information, and integrating the comparison result and the to-be-verified information to obtain a verification result.

[0052] It should be noted that by setting the cross-validation method, whether the feature region information representing the CT image data generated by multiple paths is consistent is verified, so as to improve the accuracy of obtaining the feature region information representing the CT image data. At the same time, the staff can also optimize the corresponding path according to the result of cross-validation to improve the accuracy of subsequent information generation. In the process of cross-validation method processing, the feature region information and the quantity information in the sub-information are compared in two parts to improve the efficiency of comparison. The quantity information is the number of feature region information, and the feature region information is the name of the region in the CT image data identified, such as lung nodule. The result information includes sub-information 1, sub-information 2, sub-information 3,..., and sub-information n. Similarly, the sub-feature information and the sub-quantity information correspond to the sub-information.

[0053] As shown in Figure 2As shown, the auxiliary method comprises: splitting the verification result to obtain the to-be-verified information, presetting a selection object, the selection object comprising a quantity selection, a designated selection and an auxiliary selection, determining inconsistent information items in the to-be-verified information based on the selection object to obtain replacement information items, and replacing the part of the sub-information corresponding to the replacement information items with the replacement information items to obtain optimized information.

[0054] It should be noted that when the selection object is the quantity selection, the replacement information item is obtained by extracting the information item with the most inconsistent information items in the to-be-verified information, when the selection object is the designated selection, the replacement information item is obtained by extracting the information item corresponding to the designated selection in the to-be-verified information, and when the selection object is the auxiliary selection, the feedback information is generated by integrating the to-be-verified information and fed back to the user, and the replacement information item is obtained by listening to the user feedback, the quantity selection selects the sub-feature information and the sub-quantity information with the most same quantity in the sub-information as the replacement information item, the designated selection selects the sub-feature information and the sub-quantity information generated by the designated approach as the replacement information item, and the auxiliary selection selects the corresponding sub-feature information and sub-quantity information as the replacement information item by feeding back the verification result to the administrator.

[0055] As shown in Figure 2 and Figure 5 The process of establishing a feature recognition model for analyzing different basic part features based on the storage library is as follows: data acquisition and processing: determining the recognition object of the feature recognition model to obtain a target object, traversing the storage library based on the target object to select a sub-storage library corresponding to the target object to obtain a target library, and extracting reference information in the target library to obtain training data;

[0056] It should be noted that the process of determining the recognition object of the feature recognition model to obtain the target object, i.e., determining the object analyzed by the feature recognition model, is specifically a single feature region, and the training data for model training can be obtained by finding the sub-storage library corresponding to the target object in the storage library and extracting the reference information. In actual use, the training data can be preprocessed, and the preprocessing process includes but is not limited to data cleaning, deduplication and enhanced display.

[0057] Model selection and training: selecting a deep learning model as a model base, importing the training data into the model base for the model to train to obtain an initial model;

[0058] It should be noted that the deep learning model includes but is not limited to a convolutional neural network, and the initial model is obtained by importing the training data into the model base for training.

[0059] Model adjustment and output: preset verification image and verification information, import the verification image into the initial model to obtain test information, compare the test information with the verification information to obtain the difference result, adjust and optimize the initial model based on the difference result to obtain the feature recognition model of the target object;

[0060] It should be noted that the verification image is the feature region corresponding to the target object, and the verification information is the specific information of the feature region, including but not limited to pathological information. By comparing the difference between the test information generated by the initial model and the verification information, the difference result is obtained, and the initial model is adjusted and optimized according to the difference result to obtain the feature recognition model of the target object.

[0061] Integration output: integrate all feature recognition models of the target objects to complete the establishment of feature recognition models for analyzing different feature regions.

[0062] It should be noted that a single feature recognition model corresponds to a single feature region, which can significantly improve the efficiency of analyzing the feature region and embody multi-task processing to improve the efficiency of obtaining information in the feature region.

[0063] As shown in Figure 2 The information acquisition method comprises the following steps: presetting a basic feature region, establishing a sub-repository with the basic feature region as the name, establishing a storage repository for storing the sub-repository, obtaining CT image information and feature information of the basic feature region to obtain reference information, and storing the reference information in the sub-repository.

[0064] It should be noted that the process of presetting the basic feature region is to preset the feature region that may appear in the actual use of CT image, for example, the feature region is a nodule, etc. The process of obtaining CT image information and feature information of the basic feature region to obtain reference information is to obtain the image information of the basic feature region in the processing of the past CT image information and the feature information expressing the image information. The number of basic feature regions is multiple. By storing the reference information in the sub-repository and naming it with the basic feature region, it is convenient to find the corresponding training data when training the model, and the efficiency of model training is improved.

[0065] As shown in Figure 2 The matching method comprises the following steps: splitting the analysis requirement to obtain a target feature region and an information requirement, traversing the relationship directory based on the target feature region to obtain a traversal result, selecting a feature recognition model based on the traversal result and the association relationship to obtain a selected recognition model, and adjusting the output mode of the selected recognition model based on the information requirement to obtain a target recognition model.

[0066] It should be noted that, by setting the matching method, according to the target feature area of the demand analysis, the corresponding feature recognition model is selected for the target feature area, and the output mode of the selected recognition model is adjusted based on the information demand to obtain the target recognition model, that is, the information format output by the target recognition model is adjusted according to the information format demand required to be obtained, so that the information format output meets the demand of the staff.

[0067] In the specific implementation process, different output modes can be added in the model training stage, and different information demands are corresponded by different output modes, for example, demand 1: only output the basic pathological category information of the region, example: "the region is a malignant tumor, showing irregular edges, with an area of about 12 square centimeters", demand 2: output the pathological feature parameters (such as shape, tissue type, density, etc.) of the region, example: "the region is a well-differentiated gastric tumor, with unclear boundary, density +40 HU, internal inhomogeneous structure, with rich blood vessels", demand 3: output detailed pathological related indicators and parameters (such as size, density, blood flow information, tissue type parameters, etc.), example: pathological category: malignant adenocarcinoma, size parameter: maximum radial 15mm, area 20 square millimeters, density parameter: average density +45 HU, internal inhomogeneous, blood vessel richness: high vascularization, with obvious blood flow signal, tissue characteristics: cells arranged closely, nuclei obviously heteromorphic, indicating high-grade malignant change.

[0068] The process of obtaining the feature region name and quantity information of the to-be-analyzed image data based on the verification quantity to obtain the result information is: determining the target approach for obtaining the feature region name and quantity information of the to-be-analyzed image data, obtaining the recognition accuracy of the target approach, sorting the target approaches in descending order of recognition accuracy to obtain a sorting result, selecting target approaches consistent with the verification quantity from the sorting result based on the recognition accuracy from high to low to obtain a plurality of selected approaches, obtaining sub-information reflecting the feature region name and quantity information of the to-be-analyzed image data based on the plurality of selected approaches and the to-be-analyzed image data, and integrating the plurality of sub-information to obtain the result information.

[0069] It should be noted that the process of determining the target approach for obtaining the feature region name and quantity information of the to-be-analyzed image data can obtain the target approach in the prior art according to the use demand, and the specific role of the target approach is to obtain the feature region name and quantity information of the to-be-analyzed image data.

[0070] In the specific implementation process, the target approach can be: Approach 1: Automatically extract different regions through image segmentation algorithms (such as threshold segmentation, region growing, watershed, clustering, etc.). The specific process is to pre-set the threshold, segment according to pixel intensity or texture features, and use the continuous region extraction algorithm to obtain the spatial definition of each feature region. The name (or number) and quantity of each region are counted and stored in the information structure; Approach 2: Identify by matching the features of the image region with the features in the known template or dictionary. The specific process is to build a feature dictionary (such as typical lesion features, parameters of specific structures), calculate the similarity between the regional features in the image and the template in the dictionary, identify the regions that meet the features, and name them as the corresponding region types; Approach 3: Train the model to identify different feature regions (such as using a convolutional neural network). The specific process is to use a labeled data set to train a target detection or segmentation model (such as Mask R-CNN), learn regional features, run the model to detect the target region, the model outputs the region category (name) and quantity, and extract the regional information output by the model as the "feature region name and quantity"; Approach 4: Medical experts manually or semi-automatically annotate regional information. The specific process is that experts manually annotate the region name and range according to the image characteristics, and use the knowledge base or rule system to assist in identification and statistics. Approach 5: Set rules (such as intensity range, texture parameters, geometric features) to automatically identify regions. The specific process is to apply filtering, edge detection, and connected region analysis to the image, filter regions according to parameters, count the number of regions that meet the conditions, and name the regions (such as "tumor region 1", "vascular cluster 2", etc.) according to the rules.

[0071] like Figure 6 As shown, the specification processing method includes: presetting a target format, preprocessing the CT image data to obtain preprocessed CT image data, judging whether the CT image data is consistent with the target format to obtain a judgment result, when the judgment result feedback is that the CT image data is inconsistent with the target format, horizontally scaling and vertically scaling the CT image data until it is consistent with the target format to obtain image data to be analyzed and target scaling information, and obtaining actual information based on the target scaling information and the analysis information.

[0072] It should be noted that the target format is determined by the staff, specifically the size information of the CT image data. The CT image data is set in a unified format through the set specification processing method to improve the efficiency of subsequent model processing. The actual information is obtained based on the target scaling information combined with the analysis information. That is, the actual information of the analysis information is obtained by correspondingly enlarging and reducing the unit information such as the area in the obtained analysis information with the target scaling information, such as Figure 6 As shown, Figure 6 The actual format is the original format of the CT image data, and the target scaling information is the scaling size required to adjust the actual format to the target format.

[0073] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since various modifications can be made by those skilled in the art, without departing from the spirit and scope of the application, which are defined by the appended claims and their equivalents.

Claims

1. CT image quantitative analysis system based on machine learning, including: Information acquisition module: acquires CT image data and analyzes requirements; Information processing module: Based on the CT image data, the CT image data is uniformly processed by a standard processing method to obtain the image data to be analyzed; The invention is characterized by comprising: Region acquisition module: presets the verification quantity, obtains the name and quantity information of the characteristic region of the image data to be analyzed based on the verification quantity, and obtains the result information. The result information includes several sub-information, and the number of sub-information is consistent with the verification quantity; Cross-validation module: Based on the result information, a cross-validation method is used to determine whether there are differences in the sub-information of the image data to be analyzed to obtain a verification result. Based on the verification result, an auxiliary method is used to determine the sub-information with differences to generate feedback information and optimization information; Multi-task processing module: presetting basic feature regions, establishing a repository for storing different feature regions and their feature information through an information acquisition method, establishing a feature recognition model for analyzing different feature regions based on the repository, establishing an association between the basic feature regions and the feature recognition model, and establishing a relationship directory for storing the basic feature regions, the feature recognition model, and the association relationship; Demand matching module: Based on the analysis requirements, the feature recognition model corresponding to the analysis requirements is found in the relationship directory through the matching method to obtain the target recognition model; based on the optimization information, the image data related to the analysis requirements in the image data to be analyzed are marked to obtain the target image data; the target image data is imported into the target recognition model to obtain the analysis information.

2. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The cross-validation method includes: splitting result information to obtain a plurality of sub-information, extracting feature region information from the sub-information to obtain sub-feature information, extracting feature region quantity information from the sub-information to obtain sub-quantity information, comparing whether the sub-feature information in the plurality of sub-information is consistent to obtain a first comparison result, comparing whether the sub-quantity information in the plurality of sub-information is consistent to obtain a second comparison result, obtaining a comparison result based on the first comparison result and the second comparison result, when the first comparison result and the second comparison result feedback indicate that inconsistent information exists in the plurality of sub-information, then the comparison result feedback indicates that there is a difference, extracting inconsistent information items existing in the plurality of sub-information to obtain information to be verified, and integrating the comparison result and the information to be verified to obtain a verification result.

3. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The auxiliary method includes: splitting the verification result to obtain information to be verified, presetting a selection object, the selection object including quantity selection, specified selection and auxiliary selection, determining inconsistent information items in the information to be verified based on the selection object to obtain replacement information items, and replacing the parts of the sub-information corresponding to the replacement information items with the replacement information items to obtain optimized information.

4. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The process of establishing a feature recognition model for analyzing the features of different basic parts based on the repository is as follows: Data acquisition and processing: determining the recognition object of the feature recognition model to obtain the target object, traversing the repository based on the target object to select the sub-repository corresponding to the target object to obtain the target library, and extracting the reference information in the target library to obtain the training data; Model selection and training: Select a deep learning model as the model matrix, import the training data into the model matrix for model training to obtain the initial model; Model adjustment and output: Preset verification images and verification information, import the verification images into the initial model to obtain test information, compare the test information with the verification information to obtain the difference results, and adjust and optimize the initial model based on the difference results to obtain the feature recognition model of the target object; Integrated output: Integrate the feature recognition models of all target objects to complete the establishment of feature recognition models for analysis of different feature areas.

5. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The information acquisition method includes: establishing a sub-repository with the basic feature area as the name, establishing a repository for storing the sub-repository, acquiring CT image information and feature information of the basic feature area to obtain reference information, and storing the reference information in the sub-repository.

6. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The matching method includes: splitting and analyzing requirements to obtain target feature areas and information requirements, traversing a relationship directory based on the target feature area to obtain a traversal result, selecting a feature recognition model based on the traversal result and the association relationship to obtain a selected recognition model, and adjusting the output mode of the selected recognition model based on the information requirement to obtain a target recognition model.

7. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The process of obtaining the name and quantity information of the characteristic regions of the image data to be analyzed based on the verification quantity to obtain result information is as follows: determining a path for obtaining the name and quantity information of the characteristic regions of the image data to be analyzed to obtain a target path, obtaining the recognition accuracy of the target path, sorting the target paths from high to low by the recognition accuracy to obtain a sorting result, selecting target paths consistent with the verification quantity in the sorting result based on the recognition accuracy from high to low to obtain several selected paths, obtaining sub-information reflecting the name and quantity information of the characteristic regions in the image data to be analyzed based on the several selected paths and in combination with the image data to be analyzed, and integrating the several sub-information to obtain result information.

8. The CT image quantitative analysis system based on machine learning according to claim 1, characterized in that: The specification processing method includes: presetting a target format, preprocessing CT image data to obtain preprocessed CT image data, judging whether the CT image data is consistent with the target format to obtain a judgment result, and when the judgment result feeds back that the CT image data is inconsistent with the target format, horizontally scaling and vertically scaling the CT image data until it is consistent with the target format to obtain image data to be analyzed and target scaling information, and obtaining actual information based on the target scaling information and the analysis information.

Images