Medical image analysis processing method based on AI

Abstract

The invention discloses a medical image analysis processing method based on AI, and relates to the technical field of image analysis, and the method comprises the steps: collecting multi-modal medical image data of a patient from an image device, and carrying out the preprocessing of the multi-modal medical image data; fusing the preprocessed multi-modal medical image data, extracting biological tissue features, constructing a biological tissue feature library, and marking abnormal biological tissue features in the biological tissue feature library; according to a medical image analysis task, combining with the marked biological tissue feature library to train an AI visual detection model so as to identify and classify different abnormal biological tissues; and automatically detecting an abnormal biological tissue area in the image by using an AI visual detection model. According to the method, high-precision analysis can be carried out on medical images through the AI technology, tiny abnormal changes can be detected, the severity of lesions can be evaluated and the development trend of diseases can be predicted through quantitative analysis on image features, and the risk of disease progression can be timely found and reduced when the lesions do not cause obvious symptoms.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and particularly relates to a method for analyzing and processing medical images based on AI. Background Art

[0002] Medical imaging technology is an indispensable part of modern medicine and is widely used in the early diagnosis of diseases, the formulation of treatment plans, and the monitoring of diseases. Common types of medical images include X-rays, CT (Computed Tomography), MRI (Magnetic Resonance Imaging), ultrasound, PET (Positron Emission Tomography), etc. With the continuous progress of medical technology and the aggravation of population aging, the demand for medical imaging examinations is increasing day by day. The rapid development of AI technology provides new solutions for medical image analysis and promotes the development of medical image analysis towards automation and intelligence.

[0003] For example, in the method for analyzing and processing medical images based on AI with the Chinese patent publication number: CN116823701A, the medical images are repaired and segmented, which can effectively reduce the noise components of the medical images and facilitate subsequent synchronous processing of different sub-picture parts of the medical images. It also performs recognition and comparison processing on the biological tissue characteristics of the medical images, accurately locates the lesion source, and improves the accuracy of medical image processing.

[0004] In the prior art, for the recognition and comparison processing of the biological tissue characteristics of medical images, the lesion source is quickly and accurately located, which solves the problems of long processing time and large noise interference in the processing results of digital medical images, and reduces the reliability and accuracy of digital medical image processing. However, due to the large differences in the biological tissue characteristics of different patients, there will be cases where abnormal tissues cannot be accurately identified, resulting in misdiagnosis or missed diagnosis of abnormal biological tissues. Therefore, a method for analyzing and processing medical images based on AI is proposed to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for analyzing and processing medical images based on AI to solve the problems raised in the above background art.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for analyzing and processing medical images based on AI includes the following steps:

[0008] Step 1: Collect multi-modal medical image data of a patient from an imaging device and preprocess it;

[0009] Step 2: Fusion the preprocessed multi-modal medical image data, extract biological tissue characteristics, construct a biological tissue characteristic library, and mark the abnormal biological tissue characteristics therein;

[0010] Step 3: According to the medical imaging analysis task, train the AI vision detection model by combining the labeled biological tissue feature library to identify and classify different abnormal biological tissues;

[0011] Step 4: Use the AI vision detection model to automatically detect the abnormal biological tissue areas in the image, and segment the detected abnormal biological tissues to obtain their boundaries and ranges;

[0012] Step 5: Based on the segmentation results and the detection results of the AI vision detection model, analyze the characteristics of the abnormal biological tissue areas to determine whether there is a lesion;

[0013] Step 6: Combine the analysis results of the AI vision detection model to determine the severity of the existing lesion, and then output targeted recommended measures.

[0014] A further improvement of the technical solution of the present invention is that in the step 1, the process of collecting and preprocessing multi-modal medical imaging data includes:

[0015] Connect the imaging devices (CT, MRI, PET, ultrasound, etc.) in the hospital, and obtain the imaging data of the patient through the hospital information system (PACS, i.e., Picture Archiving and Communication System), ensure the integrity and accuracy of the data. At the same time, record the patient's personal information (name, age, gender, medical record number, etc.) and the time and type of the imaging examination, assign a unique identifier to the patient, and match the patient's imaging data with the personal information;

[0016] For multiple imaging modalities (CT+MRI, PET+CT, etc.) of the same patient, register the images of different modalities spatially to ensure that the images have the same geometric relationship, and integrate them into a unified folder for subsequent processing, ensuring the spatial resolution and time series consistency of different modality data;

[0017] Preprocess the integrated multi-modal medical imaging data, including format conversion, denoising processing, and calibration processing;

[0018] Store the multi-modal medical imaging data after data quality inspection in the local server to ensure the security and accessibility of the data. Use a database management system (MySQL, MongoDB, etc.) to record the meta-information of the data. The meta-information includes patient information, examination information, imaging data format, storage location, etc., and establish a data index for quick query and retrieval of imaging data in the future.

[0019] A further improvement of the technical solution of the present invention is that in the step 2, the process of constructing the biological tissue feature library includes:

[0020] A fusion method based on deep learning, which uses a convolutional neural network to extract image features, fuses the preprocessed multi-modal medical image data of the same patient, and verifies the fused image data to check whether the fused image retains the key information of the multi-modal images, including texture information, morphological information, and functional metabolism information;

[0021] Using computer vision and image processing techniques, extract biological tissue features from the fused multi-modal image data, including morphological features, texture features, functional features, and density features. Select and optimize the extracted biological tissue features to remove redundant and irrelevant features and retain the features related to abnormal tissue detection;

[0022] Construct a biological tissue feature library, store the extracted and optimized biological tissue features into the biological tissue feature library to form a structured feature dataset. The biological tissue feature library includes the feature information of normal and abnormal biological tissues for subsequent analysis, comparison, and recognition;

[0023] Analyze the biological tissue features in the biological tissue feature library and label the abnormal biological tissue features. The labeling content includes the abnormal tissue type (such as tumor, inflammation, etc.), the location and scope of the abnormal tissue, and the clinical diagnosis information of the abnormal tissue.

[0024] A further improvement of the technical solution of the present invention is that the process of retaining the features related to abnormal tissue detection is as follows:

[0025] For morphological features, including shape, size, and edge; for texture features, including gray-level co-occurrence matrix (GLCM) features, local binary pattern (LBP) features, and wavelet transform features; for functional features, including metabolic information and blood perfusion information; for density features, including pixel value distribution and histogram features;

[0026] Select and optimize the extracted biological tissue features. Then, the features related to the abnormal tissue morphology include shape irregularity and edge blurriness; the features related to the abnormal tissue texture include texture heterogeneity and texture roughness; the features related to the abnormal tissue function include metabolic abnormality and blood perfusion abnormality; the features related to the abnormal tissue density include density inhomogeneity and density difference.

[0027] A further improvement of the technical solution of the present invention is that in the third step, the training process of the AI vision detection model includes:

[0028] Obtain the labeled feature dataset from the biological tissue feature library, and select a model architecture based on convolutional neural network in combination with the medical image analysis task. Divide the feature dataset into a training set and a test set. Among them, the feature dataset contains the feature information of multimodal medical images and the corresponding labels (normal or abnormal biological tissues), and the feature dataset covers different patients, different disease types, and different imaging modalities;

[0029] Use a deep learning framework to build a CNN model, define network structures, convolutional layers, pooling layers, fully connected layers, etc. Add regularization techniques to the model to prevent overfitting. Design a multi-task learning module (combining classification and segmentation) to make full use of feature information. Use the training set data to train the convolutional neural network model to build an AI vision detection model. During the training process, adjust the model parameters to optimize the performance, including learning rate, batch size, number of iterations, etc., and monitor the training process, evaluate the generalization ability of the model, and prevent overfitting. Use the test set data to evaluate the trained model, and calculate metrics such as the accuracy, recall rate, and F1 score of the model to measure the performance of the model. Optimize the model according to the evaluation results;

[0030] Deploy the trained AI vision detection model to the medical image analysis system, use the deployed model to extract features from new image data, and map them to the corresponding class labels to identify and classify different abnormal biological tissues.

[0031] A further improvement in the technical solution of the present invention lies in that the process of segmenting the detected abnormal biological tissue includes:

[0032] Load the already trained AI vision detection model, input the medical image data to be detected into the model. The model automatically extracts image features through the convolutional layer, pooling layer, and fully connected layer structures, and classifies each pixel in the image to determine whether it belongs to an abnormal biological tissue;

[0033] The AI vision detection model outputs a segmentation mask of the abnormal biological tissue according to the classification result, and sets a discrimination threshold to convert the probability map into a binary mask, so as to clearly distinguish abnormal tissues and normal tissues. The area with a value of 1 in the binary mask represents the abnormal biological tissue, and the area with a value of 0 represents the normal tissue. The segmentation mask is a two-dimensional or three-dimensional matrix with the same size as the input image, where the value of each pixel or voxel represents the probability that it belongs to the abnormal tissue;

[0034] Apply a series of morphological operations to process the segmentation mask. Through multiple iterative erosion and dilation operations, optimize the boundary of the segmentation mask to make it more conform to the actual shape of the abnormal tissue to optimize the segmentation result. Among them, the morphological operations include erosion, dilation, opening operation, and closing operation;

[0035] The boundary of abnormal tissue is extracted from the optimized segmentation mask using an edge detection algorithm (Canny edge detection) to obtain the boundary contour of the abnormal biological tissue. The boundary is represented as a series of coordinate points used to describe the contour of the abnormal tissue. Based on the extracted boundary contour, the boundary coordinates and range of the abnormal biological tissue are calculated, including the area or volume, centroid position, and maximum diameter. The area or volume represents the two-dimensional area or three-dimensional volume of the calculated abnormal tissue, the centroid position represents the centroid position of the calculated abnormal tissue, which is used to describe its spatial distribution, and the maximum diameter represents the longest span of the calculated abnormal tissue.

[0036] A further improvement of the technical solution of the present invention lies in that: in the fifth step, the process of determining whether there is a lesion includes:

[0037] From the segmentation mask output by the AI vision detection model, the classification probability of each pixel or region is parsed, and based on the segmentation mask, the region of the abnormal biological tissue is extracted, and the boundary coordinates and range of the obtained abnormal biological tissue are analyzed, including the area or volume, centroid position, and maximum diameter;

[0038] The morphological features, texture features, intensity features, and spatial features are fused to form a comprehensive feature description, and based on medical knowledge and clinical experience, the lesion judgment criteria for the area or volume, centroid position, and maximum diameter are set;

[0039] The fused features are compared with the lesion judgment criteria to determine whether there is a lesion in the abnormal region. If there is a lesion, the lesion is further classified into benign, malignant, or inflammatory.

[0040] A further improvement of the technical solution of the present invention lies in that: the lesion judgment criteria are as follows:

[0041] Set an area threshold A T and a volume threshold V T , if A > A T or the volume V > V T , it indicates that there is a lesion;

[0042] If the centroid position significantly deviates over time, it indicates that the lesion has dynamic changes. If the GLCM contrast of the texture feature or the standard deviation of the pixel intensity of the intensity feature significantly deviates from the normal range, it indicates that there is a lesion;

[0043] Set a maximum diameter threshold D T , if D max > D T , it indicates that there is a lesion.

[0044] A further improvement of the technical solution of the present invention lies in that: in the sixth step, the process of determining the severity of the lesion includes:

[0045] Combined with the set lesion judgment criteria, analyze the abnormal biological tissue characteristics with lesion behaviors, and according to medical knowledge and clinical experience, set the grading criteria for the severity of lesions, and divide the severity of lesions into three lesion grades: mild, moderate, and severe. Among them, the lesion characteristics of the mild lesion grade are slight, such as small area or volume, regular shape, and little impact on surrounding tissues; the lesion characteristics of the moderate lesion grade are obvious, such as moderate area or volume, irregular shape, and certain impact on surrounding tissues; the lesion characteristics of the severe lesion grade are serious, such as large area or volume, extremely irregular shape, and significant impact or invasion on surrounding tissues.

[0046] Comprehensively analyze the abnormal biological tissue characteristics with identified lesion behaviors, calculate the lesion severity evaluation coefficient, and quantify the severity of lesions.

[0047] Combined with the lesion severity evaluation coefficient and the divided lesion grades, match corresponding severity thresholds for each lesion grade.

[0048] For each lesion grade of each severity, output corresponding recommended measures.

[0049] Due to the adoption of the above technical solution, the technical progress achieved by the present invention compared with the prior art is:

[0050] 1. The present invention provides an AI-based medical image analysis and processing method. Through AI technology, high-precision analysis of medical images can be performed, detecting tiny abnormal changes. And through quantitative analysis of image features, the severity of lesions can be evaluated, and the development trend of diseases can be predicted, enabling timely detection when lesions have not yet caused obvious symptoms and reducing the risk of disease progression.

[0051] 2. The present invention provides an AI-based medical image analysis and processing method. By extracting and analyzing key features in images, misdiagnosis and missed diagnosis caused by human factors are reduced. Through deep learning algorithms, AI can quickly and accurately identify lesion areas, improving the accuracy of diagnosis. At the same time, the automated processing of AI also greatly shortens the diagnosis time, improves the doctor's work efficiency, and enables doctors to focus more on the treatment and care of patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0053] Figure 1 It is a schematic flow chart of the method of the present invention;

[0054] Figure 2 This is a schematic flow diagram for the present invention to determine whether there is a lesion. Detailed implementation manners

[0055] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, 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 some but not all of the embodiments of the present invention. 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.

[0056] Embodiment 1, as Figure 1 shown, the present invention provides an AI-based medical image analysis and processing method, including the following steps:

[0057] Step 1: Collect the multi-modal medical image data of the patient from the imaging equipment and preprocess it. Connect the hospital's imaging equipment (CT, MRI, PET, ultrasound, etc.) and obtain the patient's image data through the hospital's information system (PACS, i.e., Picture Archiving and Communication System) to ensure the integrity and accuracy of the data. At the same time, record the patient's personal information (name, age, gender, medical record number, etc.) as well as the time and type of the imaging examination, assign a unique identifier to the patient, and match the patient's image data with the personal information. For multiple imaging modalities of the same patient (CT+MRI, PET+CT, etc.), register the images of different modalities spatially to ensure that the images have the same geometric relationship and integrate them into a unified folder for subsequent processing. Ensure the spatial resolution and temporal sequence consistency of different modality data. Preprocess the integrated multi-modal medical image data, including format conversion, denoising processing, and correction processing. Among them, uniformly convert the image data output by different devices into the HDF5 format to ensure the compatibility of the data format, and use filtering algorithms (Gaussian filtering, median filtering, etc.) to remove the noise in the images, reduce artifacts and interference, and then perform geometric correction (rotation, translation, scaling) on the images to ensure the correct alignment of the images. In addition, for CT and MRI images, perform intensity correction to normalize the pixel values to a unified range (0-1 or 0-255). Furthermore, perform data quality checks on the preprocessed multi-modal medical image data, including integrity checks, consistency checks, and outlier checks. The integrity check is used to check whether the image data is complete and whether there are missing slices or data corruption. The consistency check ensures the consistency of multi-modal data in time and space and avoids errors caused by equipment differences or different examination times. The outlier check is used to check whether there are outliers in the image data (such as pixel values exceeding the normal range) for correction or elimination. Store the multi-modal medical image data after the data quality check in the local server to ensure the security and accessibility of the data. Use a database management system (MySQL, MongoDB, etc.) to record the meta-information of the data. The meta-information includes patient information, examination information, image data format, storage location, etc., and establish a data index for subsequent rapid query and retrieval of image data;

[0058] Step 2: Fuse the preprocessed multi-modal medical image data, extract biological tissue features, construct a biological tissue feature library, label the abnormal biological tissue features therein. Based on a deep learning-based fusion method, use a convolutional neural network to extract image features, fuse the preprocessed multi-modal medical image data of the same patient, and verify the fused image data to check whether the fused image retains the key information of the multi-modal images, including texture information, morphological information, and functional metabolism information. By comparing with the original images, evaluate the effect of the fused images in retaining anatomical details and functional information. Among them, the CNN automatically learns the features of different modality images, adopts a dual-branch CNN architecture to process anatomical images (CT, MRI) and functional images (PET) respectively. After extracting features through convolutional operations, use an attention mechanism to fuse the features. For multi-scale information fusion, combine local Laplacian filtering to decompose the image into multi-scale approximation images and residual images, then extract super-resolution anatomical images through the CNN and fuse them with functional images. Use computer vision and image processing techniques to extract biological tissue features from the fused multi-modal image data, including morphological features, texture features, functional features, and density features. Select and optimize the extracted biological tissue features, remove redundant and irrelevant features, and retain the features related to abnormal tissue detection. Construct a biological tissue feature library, store the extracted and optimized biological tissue features in the biological tissue feature library to form a structured feature dataset. Among them, the biological tissue feature library includes the feature information of normal and abnormal biological tissues for subsequent analysis, comparison, and identification. Analyze the biological tissue features in the biological tissue feature library and label the abnormal biological tissue features. The labeling content includes abnormal tissue types (tumors, inflammations, etc.), abnormal tissue locations and ranges, and abnormal tissue clinical diagnosis information;

[0059] In addition, retain the features related to abnormal tissue detection. The specific process is as follows:

[0060] For morphological features, including shape, size, and edge; for texture features, including gray-level co-occurrence matrix (GLCM) features, local binary pattern (LBP) features, and wavelet transform features; for functional features, including metabolic information and blood perfusion information; for density features, including pixel value distribution and histogram features. Among them, shape features describe the geometric shape of biological tissues, such as circular, oval, irregular, etc.; size features are characterized by measuring the area, volume, or diameter of the tissue; edge features reflect the changes in the tissue boundary, such as smooth, serrated, or blurred; GLCM features describe the uniformity, contrast, correlation, etc. of the texture by calculating the spatial correlation between pixel gray values; LBP features reflect the texture pattern of the local area and have a certain robustness to illumination changes; wavelet transform features extract the detailed information of the texture through multi-scale analysis; metabolic information features represent the features reflecting tissue metabolic activity extracted from PET images; blood perfusion information features represent the features reflecting tissue blood perfusion extracted from MRI or CT perfusion imaging; pixel value distribution features characterize the density of the tissue by statistically analyzing the distribution of pixel values in the image; histogram features reflect the density distribution of the tissue by calculating the gray histogram of the image;

[0061] Select and optimize the extracted biological tissue features. Then, the features related to abnormal tissue morphology include shape irregularity and edge blurriness; the features related to abnormal tissue texture include texture heterogeneity and texture roughness; the features related to abnormal tissue function include metabolic abnormalities and blood perfusion abnormalities; the features related to abnormal tissue density include density inhomogeneity and density difference. Among them, abnormal tissues usually have an irregular shape, and retaining the irregularity index in shape features helps detect abnormal tissues; the edges of abnormal tissues are relatively blurred or serrated, and retaining the blurriness or irregularity index in edge features; the texture of abnormal tissues is relatively complex and heterogeneous, and retaining the features reflecting texture heterogeneity (the contrast of GLCM, the complexity of LBP); the texture of abnormal tissues is relatively rough, and retaining the features reflecting texture roughness; abnormal tissues have a high metabolic activity, and retaining the functional features reflecting metabolic abnormalities; the blood perfusion of abnormal tissues is different from that of normal tissues, and retaining the functional features reflecting blood perfusion abnormalities; the density distribution of abnormal tissues is relatively inhomogeneous, and retaining the features reflecting density inhomogeneity; the density difference between abnormal tissues and surrounding normal tissues is large, and retaining the features reflecting density difference;

[0062] Step 3: According to the medical image analysis task, train an AI vision detection model by combining the labeled biological tissue feature library to identify and classify different abnormal biological tissues. Obtain the labeled feature dataset from the biological tissue feature library, and select a model architecture based on convolutional neural network in combination with the medical image analysis task. Divide the feature dataset into a training set and a test set. The feature dataset contains the feature information of multimodal medical images and the corresponding labels (normal or abnormal biological tissues), and the feature dataset covers different patients, different disease types, and different imaging modalities. Use a deep learning framework to build a CNN model, define the network structure, convolutional layer, pooling layer, fully connected layer, etc. Add regularization techniques to the model to prevent overfitting. Design a multi-task learning module (combining classification and segmentation) to make full use of the feature information. Use the training set data to train the convolutional neural network model and build an AI vision detection model. During the training process, adjust the model parameters to optimize the performance, including the learning rate, batch size, number of iterations, etc., and monitor the training process to evaluate the generalization ability of the model and prevent overfitting. Use the test set data to evaluate the trained model, calculate metrics such as the accuracy, recall rate, and F1 score of the model to measure the performance of the model. According to the evaluation results, optimize the model. Deploy the trained AI vision detection model to the medical image analysis system, use the deployed model to extract features from new image data, and map them to the corresponding class labels to identify and classify different abnormal biological tissues;

[0063] Step 4: Use the AI vision detection model to automatically detect abnormal biological tissue regions in the image, segment the detected abnormal biological tissues to obtain their boundaries and ranges. Load the pre-trained AI vision detection model, input the medical image data to be detected into the model. The model automatically extracts image features through convolutional layers, pooling layers, and fully connected layer structures, classifies each pixel in the image to determine whether it belongs to abnormal biological tissue. The AI vision detection model outputs the segmentation mask of abnormal biological tissues according to the classification results, sets a discrimination threshold, and converts the probability map into a binary mask to clearly distinguish abnormal tissues from normal tissues. The regions with a value of 1 in the binary mask are represented as abnormal biological tissues, and the regions with a value of 0 are represented as normal tissues. The segmentation mask is a two-dimensional or three-dimensional matrix with the same size as the input image, where the value of each pixel or voxel represents the probability that it belongs to abnormal tissues. Apply a series of morphological operations to process the segmentation mask. Through multiple iterative erosion and dilation operations, optimize the boundary of the segmentation mask to make it more conform to the actual shape of abnormal tissues to optimize the segmentation results. Among them, morphological operations include erosion, dilation, opening operation, and closing operation. Erosion is used to remove small holes or noise points in the segmentation results to make the boundary of abnormal tissues smoother. Dilation is used to fill small holes in the segmentation results to make the boundary of abnormal tissues more complete. The opening operation first erodes and then dilates, which is used to remove small objects or connecting parts. The closing operation first dilates and then erodes, which is used to fill small holes or broken parts. Use an edge detection algorithm (Canny edge detection) to extract the boundary of abnormal tissues from the optimized segmentation mask to obtain the boundary contour of abnormal biological tissues. Among them, the boundary is represented as a series of coordinate points used to describe the contour of abnormal tissues, and based on the extracted boundary contour, calculate the boundary coordinates and range of abnormal biological tissues, including area or volume, centroid position, and maximum diameter. The area or volume represents calculating the two-dimensional area or three-dimensional volume of abnormal tissues. The centroid position represents calculating the centroid position of abnormal tissues, which is used to describe its spatial distribution. The maximum diameter represents calculating the longest span of abnormal tissues;

[0064] The expression for the centroid position is:

[0065]

[0066] In the formula, is the centroid position, and respectively represent the abscissa and ordinate of the centroid, x i and y i respectively represent the abscissa and ordinate of the i-th boundary point, and n is the total number of boundary points. The centroid position reflects the central position of abnormal tissues in the image, and its value will change with the position of abnormal tissues;

[0067] The calculation expression for the two-dimensional area is:

[0068]

[0069] Wherein, A is the two-dimensional area of the abnormal tissue. The two-dimensional area formula is based on the Shoelace Formula, which is used to calculate the area of a polygon. x n+1 = x1 and y n+1 = y1, representing the closure of the polygon. The value range of A is non-negative real numbers, representing the area size of the abnormal tissue. The area A changes with the size of the abnormal tissue and is used to evaluate the severity of the lesion area;

[0070] The calculation expression of the three-dimensional area is:

[0071]

[0072] Wherein, V is the three-dimensional volume of the abnormal tissue, is the three-dimensional coordinate of the i-th boundary point. The value range of V is non-negative real numbers, representing the volume size of the abnormal tissue. The volume V changes with the size of the abnormal tissue and is used to evaluate the severity of the three-dimensional lesion area. · represents the dot product, and × represents the cross product, represents the three-dimensional coordinate vector of the j-th boundary point, represents the three-dimensional coordinate vector of the k-th boundary point. j and k are index variables used to traverse all possible point combinations. j traverses from i + 1 to n to ensure that j is always greater than i, and k traverses from j + 1 to n to ensure that k is always greater than j;

[0073] The calculation expression of the maximum diameter is:

[0074]

[0075] Wherein, D max is the maximum diameter of the abnormal tissue. The square root is used to calculate the Euclidean distance between two points. D max The value range of is non-negative real numbers. The maximum diameter reflects the longest span of the abnormal tissue, and its value changes with the size of the abnormal tissue;

[0076] Step Five: Based on the segmentation result and the detection result of the AI vision detection model, analyze the characteristics of the abnormal biological tissue area and determine whether there is a lesion;

[0077] Step Six: Combine the analysis result of the AI vision detection model, judge the severity of the existing lesion, and then output targeted recommended measures.

[0078] Example 2, such as Figure 2As shown, on the basis of Embodiment 1, the present invention provides a technical solution: Preferably, in Step 5, the process of determining whether there is a lesion includes:

[0079] From the segmentation mask output by the AI vision detection model, parse the classification probability of each pixel or region, and according to the segmentation mask, extract the region of abnormal biological tissue, analyze the boundary coordinates and range of the obtained abnormal biological tissue, including area or volume, centroid position, and maximum diameter, fuse morphological features, texture features, intensity features, and spatial features to form a comprehensive feature description, and according to medical knowledge and clinical experience, set the lesion judgment criteria for area or volume, centroid position, and maximum diameter, compare the fused features with the lesion judgment criteria to determine whether there is a lesion in the abnormal region. If there is a lesion, further classify the lesion into benign, malignant, or inflammatory. Among them, for benign lesions, the features are close to the normal range but there are slight abnormalities, such as the area or volume is slightly larger than the normal range, usually growing slowly, with a clear boundary, distinct from the surrounding tissues, and non-invasive. For malignant lesions, the features deviate significantly from the normal range, such as rapid increase in area or volume, unstable centroid position, complex texture, rapid growth, blurred boundary, possible invasion of surrounding tissues, and the ability to metastasize. For inflammation, the features are manifested as irregular boundaries, uneven intensity, and may be accompanied by an inflammatory reaction in the surrounding tissues, manifested as local redness, swelling, heat, pain, etc. On imaging, it may present as an abnormal region with unclear boundaries;

[0080] The lesion judgment criteria are as follows:

[0081] Set an area threshold A T and a volume threshold V T , if A > A T or the volume V > V T , it indicates the presence of a lesion. If the centroid position significantly deviates over time, it indicates that the lesion has dynamic changes. If the GLCM contrast of the texture feature or the pixel intensity standard deviation of the intensity feature significantly deviates from the normal range, it indicates the presence of a lesion. Set a maximum diameter threshold D T , if D max > D T , it indicates the presence of a lesion;

[0082] In Step 6, the process of determining the severity of the lesion includes:

[0083] Combined with the set lesion judgment criteria, analyze the abnormal biological tissue characteristics with lesion behaviors, and according to medical knowledge and clinical experience, set the grading criteria for the severity of the lesion. The severity of the lesion is divided into three grades: mild, moderate, and severe. Among them, the lesion characteristics of the mild lesion grade are slight, such as small area or volume, regular shape, and little impact on the surrounding tissues. The lesion characteristics of the moderate lesion grade are obvious, such as moderate area or volume, irregular shape, and certain impact on the surrounding tissues. The lesion characteristics of the severe lesion grade are severe, such as large area or volume, extremely irregular shape, and significant impact or invasion on the surrounding tissues. Comprehensively analyze the abnormal biological tissue characteristics with lesion behaviors identified, calculate the lesion severity evaluation coefficient, and quantify the severity of the lesion. Combine the lesion severity evaluation coefficient and the divided lesion grades to match corresponding severity thresholds for each lesion grade. For each lesion grade with different severities, output corresponding recommended measures. For the mild lesion grade, it is recommended to conduct imaging reexaminations every 3 - 6 months to monitor the development of the lesion, adjust living habits such as diet and work and rest to promote health, and if necessary, carry out drug treatment to control the progression of the lesion. Mild lesions usually do not cause serious impacts on the patient's life and health, but it is necessary to closely monitor their development to promptly detect and handle potential risks. For the moderate lesion grade, immediately conduct detailed medical examinations to clarify the nature and cause of the lesion. According to the examination results, formulate personalized treatment plans such as surgery, radiotherapy, chemotherapy, etc., strengthen the patient's life care and nutritional support, and improve the patient's immunity. Moderate lesions may cause certain impacts on the patient's life and health, so it is necessary to promptly conduct detailed medical examinations to clarify the nature and cause of the lesion and formulate corresponding treatment plans. For the severe lesion grade, immediately initiate an emergency treatment procedure, conduct multidisciplinary consultations, formulate a comprehensive treatment plan, closely monitor the changes in the patient's vital signs, promptly adjust the treatment plan, and provide the patient with comprehensive life care and psychological support to improve the patient's quality of life. Severe lesions may pose a serious threat to the patient's life and health, so it is necessary to immediately initiate an emergency treatment procedure, conduct multidisciplinary consultations, formulate a comprehensive treatment plan. At the same time, closely monitor the changes in the patient's vital signs, promptly adjust the treatment plan, and provide the patient with comprehensive life care and psychological support;

[0084] The expression of the lesion severity evaluation coefficient is:

[0085]

[0086] In the formula, SS is the lesion severity evaluation coefficient, and the higher the value, the more severe the lesion. A is the area or volume of the abnormal tissue, and A0 is the reference value of the area or volume, which is used for normalization and represents the maximum area or volume of the normal tissue. D maxis the maximum diameter of the abnormal tissue, D0 is the reference value of the maximum diameter, which is used for normalization and represents the maximum diameter of the normal tissue, ΔC is the change in the centroid position, that is, the offset compared with the previous examination, which is used to evaluate the dynamic change of the lesion, and α, β, and γ are weight coefficients, which are used to adjust the contribution of each feature to the severity evaluation. The value range of SS is from 0 to 1, where 0 indicates no lesion and 1 indicates the most severe lesion. When A, D max or ΔC increases, the SS value increases, indicating an increase in the severity of the lesion;

[0087] Multiple lesion grades correspond to multiple severity thresholds. Among them, the severity thresholds include an upper threshold and a lower threshold;

[0088] Multiple lesion grades and multiple severity thresholds satisfy the following relationship:

[0089] Mild lesion grade 0 < SS ≤ SS L ;

[0090] Moderate lesion grade SS L < SS ≤ SS M ;

[0091] Severe lesion grade SS M < SS < 1;

[0092] Among them, SS is the lesion severity evaluation coefficient, and SS L is the lower threshold corresponding to the moderate lesion grade and the upper threshold corresponding to the mild lesion grade, and SS M is the lower threshold corresponding to the severe lesion grade and the upper threshold corresponding to the moderate lesion grade, SS L = 0.33, SS M = 0.67.

[0093] As mentioned above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An AI-based medical image analysis and processing method, characterized in that, It includes the following steps: Step 1: Collect the multimodal medical image data of the patient from the imaging device and preprocess it; Step 2: Fuse the preprocessed multimodal medical image data, extract the biological tissue features, construct a biological tissue feature library, and label the abnormal biological tissue features therein; Step 3: According to the medical image analysis task, train an AI vision detection model in combination with the labeled biological tissue feature library to identify and classify different abnormal biological tissues; Step 4: Use the AI vision detection model to automatically detect the abnormal biological tissue regions in the image, and segment the detected abnormal biological tissues to obtain their boundaries and ranges; Step 5: Based on the segmentation results and the detection results of the AI vision detection model, analyze the characteristics of the abnormal biological tissue regions and determine whether there is a lesion; Step 6: Combine the analysis results of the AI vision detection model, judge the severity of the existing lesion, and then output targeted recommended measures.

2. The method for analyzing and processing medical images based on AI according to claim 1, characterized in that: In the said Step 1, the process of collecting and preprocessing the multimodal medical image data includes: Connect to the imaging device in the hospital, and obtain the patient's image data through the hospital's information system. At the same time, record the patient's personal information, as well as the time and type of the imaging examination, assign a unique identifier to the patient, and match the patient's image data with the personal information; For multiple image modalities of the same patient, register the images of different modalities spatially and integrate them into a unified folder; Preprocess the integrated multimodal medical image data, including format conversion, denoising processing, and calibration processing; Store the multimodal medical image data after data quality inspection in the local server, use a database management system to record the metadata of the data, and establish a data index.

3. The method for analyzing and processing medical images based on AI according to claim 2, wherein: In the said Step 2, the process of constructing the biological tissue feature library includes: Based on the deep learning-based fusion method, use a convolutional neural network to extract image features, fuse the preprocessed multimodal medical image data of the same patient, and verify the fused image data to verify whether the fused image retains the key information of the multimodal images, including texture information, morphological information, and functional metabolism information; Use computer vision and image processing techniques to extract biological tissue features from the fused multimodal image data, including morphological features, texture features, functional features, and density features, select and optimize the extracted biological tissue features, remove redundant and irrelevant features, and retain the features related to abnormal tissue detection; Construct a biological tissue feature library, store the extracted and optimized biological tissue features in the biological tissue feature library to form a structured feature data set. Among them, the biological tissue feature library includes the feature information of normal and abnormal biological tissues; Analyze the biological tissue features in the biological tissue feature library and label the abnormal biological tissue features therein. The labeling content includes the abnormal tissue type, the location and range of the abnormal tissue, and the clinical diagnosis information of the abnormal tissue.

4. The method for AI-based medical image analysis and processing according to claim 3, wherein: The specific process of retaining the features related to abnormal tissue detection is: For morphological features, including shape, size, and edges, for texture features, including gray-level co-occurrence matrix features, local binary pattern features, and wavelet transform features, for functional features, including metabolic information and blood perfusion information, and for density features, including pixel value distribution and histogram features; Select and optimize the extracted biological tissue features. Then, the features related to abnormal tissue morphology include shape irregularity and edge blurriness, the features related to abnormal tissue texture include texture heterogeneity and texture roughness, the features related to abnormal tissue function include metabolic abnormalities and blood perfusion abnormalities, and the features related to abnormal tissue density include density inhomogeneity and density difference.

5. The method for analyzing and processing medical images based on AI according to claim 4, characterized in that: In the third step, the training process of the AI vision detection model includes: Obtain the labeled feature dataset from the biological tissue feature library, and select a model architecture based on a convolutional neural network in combination with the medical image analysis task. Divide the feature dataset into a training set and a test set. Among them, the feature dataset contains the feature information of multimodal medical images and the corresponding labels, and the feature dataset covers different patients, different disease types, and different imaging modalities; Use a deep learning framework to build a CNN model, use the training set data to train the convolutional neural network model, build an AI vision detection model. During the training process, adjust the model parameters to optimize the performance, use the test set data to evaluate the trained model, and optimize the model according to the evaluation results; Deploy the trained AI vision detection model to the medical image analysis system, use the deployed model to extract features from new image data, and map them to the corresponding class labels to identify and classify different abnormal biological tissues.

6. The method for analyzing and processing medical images based on AI according to claim 5, characterized in that: In the fourth step, the process of segmenting the detected abnormal biological tissue includes: Load the trained AI vision detection model, input the medical image data to be detected into the model. The model automatically extracts image features through the convolutional layer, pooling layer, and fully connected layer structures, and classifies each pixel in the image to determine whether it belongs to abnormal biological tissue; The AI vision detection model outputs the segmentation mask of the abnormal biological tissue according to the classification results, and sets a discrimination threshold to convert the probability map into a binary mask, so as to clearly distinguish abnormal tissue from normal tissue. The area with a value of 1 in the binary mask represents abnormal biological tissue, and the area with a value of 0 represents normal tissue; Apply a series of morphological operations to process the segmentation mask. Through multiple iterations of erosion and dilation operations, optimize the boundary of the segmentation mask to make it more conform to the actual shape of the abnormal tissue to optimize the segmentation result. Among them, the morphological operations include erosion, dilation, opening operation, and closing operation; Use an edge detection algorithm to extract the boundary of the abnormal tissue from the optimized segmentation mask to obtain the boundary contour of the abnormal biological tissue, and calculate the boundary coordinates and range of the abnormal biological tissue according to the extracted boundary contour, including area or volume, centroid position, and maximum diameter.

7. The method for AI-based medical image analysis and processing according to claim 6, characterized in that: In the fifth step, the process of determining whether there is a lesion includes: From the segmentation mask output by the AI vision detection model, parse the classification probability of each pixel or region, and based on the segmentation mask, extract the region of abnormal biological tissue, and analyze the boundary coordinates and range of the obtained abnormal biological tissue, including area or volume, centroid position, and maximum diameter; Fuse morphological features, texture features, intensity features, and spatial features to form a comprehensive feature description, and based on medical knowledge and clinical experience, set the lesion judgment criteria for area or volume, centroid position, and maximum diameter; Compare the fused features with the lesion judgment criteria to determine whether there is a lesion in the abnormal region. If there is a lesion, further classify the lesion into benign, malignant, or inflammatory.

8. The method for analyzing and processing medical images based on AI according to claim 7, wherein: The lesion judgment criteria are as follows: Set the area threshold A T and the volume threshold V T , if the area A > A T or the volume V > V T , then prompt that there is a lesion; If the centroid position significantly deviates over time, it indicates dynamic changes in the lesion. If the GLCM contrast of the texture feature or the pixel intensity standard deviation of the intensity feature significantly deviates from the normal range, it indicates the presence of a lesion; Set the maximum diameter threshold D T , if D max > D T , then prompt the existence of a lesion.

9. The method for analyzing and processing medical images based on AI according to claim 8, wherein: In step six, the process of judging the severity of the lesion includes: Combined with the set lesion judgment criteria, analyze the characteristics of abnormal biological tissue with lesion behavior, and based on medical knowledge and clinical experience, set the grading criteria for the severity of the lesion, and divide the severity of the lesion into three lesion grades: mild, moderate, and severe; Comprehensively analyze the characteristics of abnormal biological tissue with lesion behavior identified, calculate the lesion severity evaluation coefficient, and quantify the severity of the lesion; Combined with the lesion severity evaluation coefficient and the divided lesion grades, match the corresponding severity thresholds for each lesion grade; For each severity level of the lesion grade, output the corresponding recommended measures.

Images