Kidney transplantation rejection reaction subtype classification system based on artificial intelligence

Abstract

The invention discloses a kidney transplantation rejection reaction subtype classification system based on artificial intelligence, which relates to the field of kidney transplantation and comprises a data collection module, a feature aggregation and classification module, a feature extraction module and a model evaluation module. The system is used for extracting pathological image feature vectors and clinical feature vectors according to the collected pathological image data and the clinical feature number, and then splicing and classifying the pathological image feature vectors and the clinical feature vectors to obtain a subtype classification result of the kidney transplantation rejection reaction; and a classification result is evaluated through a confusion matrix, a receiver operation characteristic curve ROC and an area under the curve AUC so as to ensure the accuracy and reliability of the result. According to the invention, the accuracy and repeatability of diagnosis can be improved, the workload of pathologists can be reduced, multi-modal information can be more fully integrated, and the comprehensiveness and precision of diagnosis can be improved.

Description

Technical Field

[0001] The present invention relates to the field of kidney transplantation, and in particular to an artificial intelligence-based classification system for kidney transplant rejection subtypes. Background Art

[0002] In the field of kidney transplantation, accurate diagnosis of rejection is crucial for ensuring the long-term survival of the transplanted kidney. Currently, the clinical diagnosis of kidney transplant rejection relies primarily on manual histopathological evaluation. Specifically, pathologists observe H&E-stained sections of renal biopsy tissue under a microscope and, based on specific histological features such as tubular damage, interstitial inflammation, and vascular lesions, combine the Banff classification criteria to diagnose and subtype rejection. In addition, clinical indicators such as blood creatinine levels and postoperative time are also considered to assist in diagnosis.

[0003] However, the existing technical solutions have the following disadvantages:

[0004] High Subjectivity: Manual histopathology evaluation relies heavily on the subjective judgment of pathologists. Diagnostic results for the same biopsy sample may differ between different pathologists, leading to poor diagnostic reproducibility. This variability may stem from differences in pathologist experience, understanding, and application of the Banff classification criteria.

[0005] High workload: The evaluation process requires pathologists to spend a lot of time and energy to carefully observe and analyze the histological characteristics of each biopsy sample. For high-throughput sample processing, the workload is heavy and the diagnostic accuracy is easily affected by factors such as fatigue.

[0006] Insufficient information integration: While some clinical indicators are used to aid diagnosis, existing methods often fail to fully integrate multimodal information, such as biopsy pathology images and clinical features, to achieve a more comprehensive and accurate diagnosis. For example, in some complex cases, it may be difficult to accurately determine the subtype of rejection based solely on pathology images or a single clinical indicator. However, comprehensive analysis of multimodal information is expected to improve diagnostic accuracy. Summary of the Invention

[0007] In order to solve the above problems, the purpose of the present invention is to provide an artificial intelligence-based kidney transplant rejection reaction subtype classification technology, which aims to realize the automated and objective detection and subtype classification of kidney transplant rejection reaction by integrating multimodal information such as biopsy pathological images and clinical characteristics of kidney transplant recipients.

[0008] To achieve the above technical objectives, the present application provides an artificial intelligence-based kidney transplant rejection subtype classification system, comprising:

[0009] A data collection module is used to collect pathological image data and clinical feature data;

[0010] A feature extraction module is used to extract pathological image feature vectors and clinical feature vectors based on pathological image data and clinical feature numbers;

[0011] The feature aggregation and classification module is used to concatenate the pathological image feature vector and the clinical feature vector, and then use SVM or DNN for classification to output the subtype classification results of kidney transplant rejection;

[0012] The model evaluation module is used to evaluate the classification results through confusion matrix, receiver operating characteristic curve (ROC) and area under the curve (AUC) to ensure the accuracy and reliability of the results.

[0013] Preferably, the feature extraction module includes:

[0014] Pathology image feature extraction: Pathology image feature vectors are extracted using the pathology image data using an image feature extraction model. The image feature extraction model includes one of the following: a ResNet50 pre-trained CNN model, InceptionV3, EfficientNet-B5, or Vision Transformer.

[0015] Clinical feature extraction: clinical feature vectors are extracted using clinical feature numbers by using a feature extraction algorithm, wherein the feature extraction algorithm includes one of a random forest algorithm, a gradient boosted tree (GBDT), a support vector machine (SVM), or a multilayer perceptron (MLP).

[0016] Preferably, the feature aggregation and classification module also performs classification through a deep neural network DNN, a convolutional neural network CNN or a recurrent neural network RNN.

[0017] Preferably, the feature extraction module is further used to form a pathological image feature vector after performing feature aggregation using one of average pooling, maximum pooling, attention mechanism or clustering-based methods; and extract the clinical feature vector through a random forest algorithm.

[0018] Preferably, the feature aggregation and classification module is further used to perform classification using a support vector machine (SVM), logistic regression (LR) or decision tree (DT).

[0019] Preferably, the feature extraction module is further used to extract pathological image features using multiple instance learning and convolutional neural networks, and to extract clinical features using a random forest algorithm.

[0020] Preferably, the feature aggregation and classification module is also used to perform classification through a deep neural network DNN, a convolutional neural network CNN or a recurrent neural network RNN ​​after splicing and fusion through weighted fusion, fusion layer or attention mechanism.

[0021] Preferably, the feature aggregation and classification module is also used to perform classification through support vector machine SVM, logistic regression LR or decision tree DT after splicing and fusion through weighted fusion, fusion layer or attention mechanism.

[0022] Preferably, the model evaluation module is further used to evaluate the classification results through F1 score, precision-recall curve or Kappa statistic to ensure the accuracy and reliability of the results.

[0023] Preferably, the system also includes a data preprocessing module for performing image segmentation and normalization processing on pathological image data, or image enhancement, denoising, contrast adjustment or color correction processing; and for performing standardization processing, normalization processing, discretization processing or feature selection processing on clinical image data.

[0024] The present invention discloses the following technical effects:

[0025] The present invention integrates multimodal information and utilizes advanced feature extraction and classification methods to achieve automated and objective detection and subtype classification of kidney transplant rejection reactions, ensuring that the technical solution of the present invention is innovative and practical in many aspects, and provides a new and efficient solution for diagnosis and prognostic assessment in the field of kidney transplantation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 is a system architecture diagram of the present invention;

[0028] Figure 2 is an example of a confusion matrix according to the present invention;

[0029] Figure 3 This is an example of the ROC curve described in the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of this application.

[0031] like Figure 1-Figure 3 As shown, the present invention provides an artificial intelligence-based renal transplant rejection subtype classification system. By using deep learning technology to extract features from whole-slice images (WSIs) of renal transplant biopsies, and combining feature extraction algorithms to extract features from clinical features, the two features are aggregated and globally classified, achieving automated, objective detection and subtype classification of renal transplant rejection. The system designed by the present invention specifically includes the following contents:

[0032] 1. Data collection module:

[0033] (1) Pathological image data: H&E-stained whole-slide images (WSIs) of kidney transplant biopsies from a single hospital between January 2015 and December 2023 were collected.

[0034] (2) Clinical characteristics data: Clinical information such as age at biopsy, postoperative time, serum creatinine at biopsy, changes in serum creatinine, DSA, C4d, and absolute lymphocyte count (LYMPH) were collected.

[0035] (3) Sample classification: Biopsy samples were classified into T cell-mediated rejection (TCMR), antibody-mediated rejection (ABMR), mixed rejection (MIXR), and other lesions (Others) based on the consensus reached by two experienced transplant pathologists.

[0036] 2. Feature extraction module:

[0037] (1) Pathological image feature extraction:

[0038] A. Image Preprocessing: WSI images are preprocessed, including image segmentation and normalization, to improve image quality and reduce noise. In addition to image segmentation and normalization, other preprocessing methods can also be used, such as image enhancement, denoising, contrast adjustment, or color correction. These preprocessing methods can further improve image quality, reduce noise interference, and enhance the accuracy of feature extraction.

[0039] B. Deep Learning: The WSIs image is divided into multiple small patches. Features of each patch are extracted using a pre-trained image feature extraction model (CNN, such as ResNet50). The image feature extraction model includes: using ResNet50 as a pre-trained CNN model, InceptionV3, EfficientNet-B5, and Vision Transformer (ViT).

[0040] C. Feature aggregation: Use a multi-instance pooling layer (such as average pooling) to aggregate the features of all small blocks into a global feature vector. In addition to using average pooling for multi-instance learning, you can also use maximum pooling, attention mechanism or clustering-based methods for feature aggregation.

[0041] (2) Clinical feature extraction:

[0042] A. Feature Standardization: Clinical features are standardized to eliminate dimensional differences between them. In addition to standardization, other preprocessing methods can be used, such as normalization, discretization, or feature selection. These preprocessing methods can further optimize clinical features and improve their representativeness and effectiveness.

[0043] B. Feature extraction algorithm: Use a feature extraction algorithm to extract clinical features and generate a clinical feature vector, where the feature extraction algorithm includes: a random forest algorithm, a gradient boosted tree (GBDT), a support vector machine (SVM), or a neural network (such as a multilayer perceptron MLP).

[0044] 3. Feature aggregation and classification module:

[0045] (1) Feature splicing: The pathological image feature vector and the clinical feature vector are spliced ​​to form a comprehensive feature vector. In addition to simply splicing the pathological image feature vector and the clinical feature vector, more complex fusion strategies can also be used, such as weighted fusion, fusion layers (such as fully connected layers or convolutional layers), or attention mechanisms. These fusion strategies can more effectively integrate multimodal information and improve the comprehensiveness and representativeness of features.

[0046] (2) Classifier: Use support vector machine (SVM) or deep neural network (DNN) to classify the comprehensive feature vector and output the subtype classification results of kidney transplant rejection reaction. Among them, the classifier can use different deep learning models for classification, such as support vector machine (SVM), logistic regression (LR) or decision tree (DT).

[0047] 4. Model evaluation module:

[0048] (1) Confusion matrix: Use the confusion matrix to evaluate the classification accuracy, recall rate, precision rate and other indicators of the model.

[0049] (2) ROC curve and AUC: The classification performance of the model was evaluated using the receiver operating characteristic curve (ROC) and the area under the curve (AUC).

[0050] The working principle of the system designed by the present invention is as follows:

[0051] Data collection: H&E-stained WSIs and clinical characteristics of renal transplant biopsies were obtained from the hospital information system. Two experienced transplant pathologists classified the biopsy specimens to ensure data accuracy and consistency.

[0052] Feature extraction:

[0053] Pathology image feature extraction:

[0054] Image preprocessing: Segment and normalize the WSIs image to generate multiple patches.

[0055] Deep learning: Each small block is extracted with a pre-trained CNN (such as ResNet50) to generate a feature vector.

[0056] Feature aggregation: Use a multi-instance pooling layer (such as average pooling) to aggregate the features of all small patches into a global feature vector.

[0057] Clinical feature extraction:

[0058] Feature standardization: Standardize clinical features to eliminate dimensional differences.

[0059] Feature extraction algorithm: Use the feature extraction algorithm to extract clinical features and generate clinical feature vectors.

[0060] Feature aggregation and classification:

[0061] Feature splicing: Splice the pathological image feature vector and the clinical feature vector to form a comprehensive feature vector.

[0062] Classifier: Use SVM or DNN to classify the comprehensive feature vector and output the subtype classification results of kidney transplant rejection.

[0063] Model Evaluation:

[0064] Confusion matrix: The confusion matrix is ​​used to evaluate the classification accuracy, recall rate, precision rate and other indicators of the model, such as Figure 2 As shown in the figure, the comparison between the model classification results and the actual labels is shown, and indicators such as classification accuracy, recall rate, and precision are evaluated.

[0065] ROC curve and AUC: The classification performance of the model is evaluated by ROC curve and AUC to ensure the reliability and effectiveness of the model, such as Figure 3 As shown in the figure, the classification performance of the model is demonstrated, and the reliability of the model is evaluated by the AUC value.

[0066] Compared with the existing technology, the AI-based kidney transplant rejection subtype classification system disclosed in the present invention has the following significant advantages and beneficial effects:

[0067] 1. Improve diagnostic accuracy:

[0068] (1) Multimodal information integration: This invention not only utilizes H&E-stained whole-slice images (WSIs), but also incorporates clinical features such as age at biopsy, postoperative time, serum creatinine at biopsy, changes in serum creatinine, DSA, C4d, and absolute lymphocyte count (LYMPH). This integration of multimodal information can more comprehensively reflect the pathological and physiological status of kidney transplant recipients, thereby improving diagnostic accuracy.

[0069] (2) Deep learning and feature extraction algorithms: Multiple instance learning and image feature extraction models are used to extract features from pathological images, and feature extraction algorithms are used to extract features from clinical features. These advanced algorithms can extract and integrate features more effectively, further improving the accuracy of diagnosis.

[0070] 2. Improve diagnostic efficiency:

[0071] (1) Automated Assessment: The system of the present invention can automatically process and analyze large amounts of pathological images and clinical data, significantly reducing the workload of pathologists. The time required to complete each diagnostic task has been reduced from several hours to less than 30 seconds, greatly improving diagnostic efficiency.

[0072] (2) High-throughput processing: The automated system can process multiple samples simultaneously, which is suitable for high-throughput clinical environments and can quickly screen and classify a large number of kidney transplant biopsy samples.

[0073] 3. Improve repeatability:

[0074] (1) Objectivity: The diagnostic results of the AI ​​system are not affected by human factors and are highly objective. This makes the diagnostic results more reproducible among different pathologists and reduces misdiagnosis and missed diagnosis caused by subjective judgment differences.

[0075] (2) Standardized process: The system of the present invention adopts standardized data processing and feature extraction processes to ensure the consistency and reliability of each diagnostic process, further improving the repeatability of the diagnostic results.

[0076] 4. Scalability and flexibility:

[0077] (1) Multi-center application: The system of the present invention can be easily expanded to multiple centers and applied to pathology data from different hospitals and laboratories. This helps to improve the universality and application scope of the system.

[0078] (2) Continuous optimization: By continuously updating and optimizing the model, the present invention can adapt to new pathological characteristics and clinical data, maintaining the advancement and accuracy of the system.

[0079] In summary, the AI-based renal transplant rejection subtype classification system and method of the present invention has significant advantages in improving diagnostic accuracy, efficiency, and repeatability, and provides a new and efficient solution for the diagnosis and evaluation of renal transplantation.

[0080] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0081] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0082] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An artificial intelligence-based classification system for kidney transplant rejection subtypes, characterized by: include: A data collection module is used to collect pathological image data and clinical feature data; A feature extraction module, configured to extract a pathological image feature vector and a clinical feature vector based on the pathological image data and the clinical feature number; A feature aggregation and classification module is used to concatenate the pathological image feature vector and the clinical feature vector, perform classification using SVM or DNN, and output a subtype classification result of kidney transplant rejection reaction; The model evaluation module is used to evaluate the classification results through confusion matrix, receiver operating characteristic curve (ROC) and area under the curve (AUC) to ensure the accuracy and reliability of the results.

2. The artificial intelligence-based renal transplant rejection subtype classification system according to claim 1, characterized in that: The feature extraction module includes: Pathology image feature extraction: extracting a pathology image feature vector using the pathology image data by using an image feature extraction model, wherein the image feature extraction model includes: using ResNet50 as a pre-trained CNN model, InceptionV3, EfficientNet-B5, and Vision Transformer; Clinical feature extraction: Utilizing the clinical feature number, the clinical feature vector is extracted by using a feature extraction algorithm, wherein the feature extraction algorithm includes one of a random forest algorithm, a gradient boosting tree (GBDT), a support vector machine (SVM), or a multilayer perceptron (MLP).

3. The artificial intelligence-based renal transplant rejection subtype classification system according to claim 2, characterized in that: The feature aggregation and classification module also performs classification through a deep neural network DNN, a convolutional neural network CNN or a recurrent neural network RNN.

4. The artificial intelligence-based renal transplant rejection subtype classification system according to claim 1, characterized in that: The feature extraction module is further used to form the pathological image feature vector after performing feature aggregation using one of average pooling, maximum pooling, attention mechanism or clustering-based methods; and extract the clinical feature vector through a random forest algorithm.

5. The artificial intelligence-based renal transplant rejection subtype classification system according to claim 4, characterized in that: The feature aggregation and classification module is further used to perform classification using a support vector machine (SVM), a logistic regression (LR), or a decision tree (DT).

6. The artificial intelligence-based kidney transplant rejection subtype classification system according to claim 1, characterized in that: The feature extraction module is also used to extract pathological image features using multiple instance learning and convolutional neural networks, and to extract clinical features using a random forest algorithm.

7. The artificial intelligence-based kidney transplant rejection subtype classification system according to claim 6, characterized in that: The feature aggregation and classification module is also used to perform classification through a deep neural network DNN, a convolutional neural network CNN or a recurrent neural network RNN ​​after splicing and fusion through weighted fusion, fusion layer or attention mechanism.

8. The artificial intelligence-based kidney transplant rejection subtype classification system according to claim 6, characterized in that: The feature aggregation and classification module is also used to perform classification through support vector machine SVM, logistic regression LR or decision tree DT after splicing and fusion through weighted fusion, fusion layer or attention mechanism.

9. The artificial intelligence-based kidney transplant rejection subtype classification system according to claim 8, characterized in that: The model evaluation module is further used to evaluate the classification results through F1 score, precision-recall curve or Kappa statistic to ensure the accuracy and reliability of the results.

10. The artificial intelligence-based kidney transplant rejection subtype classification system according to claim 1, characterized in that: The system also includes a data preprocessing module for performing image segmentation and normalization processing, or image enhancement, denoising, contrast adjustment or color correction processing on the pathological image data; and for performing standardization processing, normalization processing, discretization processing or feature selection processing on the clinical image data.

Images