Computer-aided medical diagnosis system and method

Abstract

The invention discloses a computer-aided medical diagnosis system and method, and relates to the technical field of computer-aided medical diagnosis systems.The computer-aided medical diagnosis method comprises the steps that patient medical data from different sources are cleaned, denoised and standardized through a data collecting and preprocessing method, and a high-quality data set is obtained; constructing a diagnosis model based on the high-quality data set and a deep learning algorithm, inputting the high-quality data set into the diagnosis model, and outputting a diagnosis result; integrating the data used by the diagnosis result by adopting a multi-modal data fusion method, integrating the data from different sources into a unified framework for analysis, and obtaining a data set after multi-modal fusion processing; based on the data set after the multi-modal fusion processing, optimizing the diagnosis model by adopting a cross validation technology, and inputting the data set after the multi-modal fusion processing into the optimized diagnosis model to obtain an optimized diagnosis result; and based on the optimized diagnosis result, generating a diagnosis report by using an explanatory algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of computer-aided medical diagnosis systems, and in particular to a computer-aided medical diagnosis system and method. Background Art

[0002] Computer-aided medical diagnosis system is a technology that uses advanced computing technology and artificial intelligence algorithms to support medical diagnosis. It provides more accurate and personalized diagnostic recommendations by integrating and analyzing medical data from multiple sources (such as electronic health records, medical images, and genomic data).

[0003] In the field of computer-assisted medical diagnosis systems, in existing medical diagnosis systems, data from different resources usually have different formats, noise levels and data integrity issues, and many existing deep learning models perform well on the training set, but when faced with new patients or rare cases, they are prone to overfitting, resulting in insufficient generalization ability. At the same time, existing computer-assisted medical diagnosis systems often lack user-friendly interfaces, making it difficult for doctors to easily view and understand the diagnosis results and the basis behind them. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a computer-aided medical diagnosis method to solve the problem in existing medical diagnosis systems that data from different sources usually have different formats, noise levels and data integrity.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a computer-aided medical diagnosis method, comprising:

[0008] Data collection and preprocessing methods are used to clean, denoise, and standardize patient medical data from different sources to obtain high-quality datasets;

[0009] Build a diagnostic model based on high-quality data sets and deep learning algorithms, input high-quality data sets into the diagnostic model, and output diagnostic results;

[0010] The multimodal data fusion method is used to integrate the data used for the diagnosis results, integrating data from different sources into a unified framework for analysis to obtain a multimodal fusion processed data set;

[0011] Based on the data set processed by multimodal fusion, the cross-validation technology is used to optimize the diagnostic model, and the data set processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic results;

[0012] Based on the optimized diagnostic results, an explanatory algorithm is used to generate a diagnostic report, and visualization technology is used to display the key information in the diagnostic report to obtain a visual diagnostic result.

[0013] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, the data collection and preprocessing method is used to clean, denoise, and standardize patient medical data from different sources to obtain a high-quality data set. The specific steps are as follows:

[0014] Collect raw data from electronic health records, medical imaging, and genomic data;

[0015] Use natural language processing (NLP) to segment text data, remove stop words, and perform stemming operations. Calculate the importance score of each word using the TF-IDF algorithm to achieve data cleaning.

[0016] Apply Gaussian filtering to the image data to smooth the image and reduce noise;

[0017] The Z-Score normalization method was used to transform genomic data into a standard normal distribution and obtain a high-quality dataset.

[0018] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, the diagnostic model is constructed based on a high-quality data set and a deep learning algorithm, and the high-quality data set is input into the diagnostic model to output a diagnostic result. The specific steps are as follows:

[0019] A diagnostic model is constructed based on high-quality data sets and convolutional neural networks (CNNs), which can be expressed as:

[0020]

[0021] Among them, L represents the total loss, y i and Represent the true label and predicted value of the i-th sample respectively, N is the total number of samples, λ is the regularization coefficient used to control the complexity of the model, Ω represents the weight space, Indicates the absolute value of the weight change rate;

[0022] Input high-quality data sets into the diagnostic model for training, the expression is:

[0023]

[0024] Where BN(x) represents the normalized output, x is the input data, E[x] and Var[x] are the mean and variance of the input data, respectively. It is a small constant for numerical stability, and γ and β are learnable parameters used to scale and translate the normalized data.

[0025] Input a high-quality data set into the trained diagnostic model and output the diagnostic result. The expression is:

[0026]

[0027] Among them, P(y|x) represents the probability that the model predicts the output as category y given the input data x, x is the input data, x i represents the i-th eigenvalue of the input data, f(x i ) is a nonlinear activation function, w i is the weight parameter associated with the i-th feature, σ is the Sigmoid function;

[0028] The preliminary diagnostic results obtained will serve as the basis for subsequent multimodal data fusion to further improve the comprehensiveness and accuracy of the diagnosis.

[0029] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, the multimodal data fusion method is used to integrate the data used for the diagnosis results, integrating the data from different sources into a unified framework for analysis to obtain a multimodal fusion data set. The specific steps are as follows:

[0030] Define a multimodal fusion function to integrate data from different sources;

[0031] Set up a fusion model based on the attention mechanism to dynamically adjust the importance weights of different modal data;

[0032] The importance weight of each modality is calculated using the attention mechanism, and the expression is:

[0033]

[0034] Among them, α i is the importance weight of the i-th mode, e i is the scoring function output of the i-th mode;

[0035] F=α ehr ·E ehr +α img ·E img +α gen ·E gen ;

[0036] Among them, F is the final fusion feature vector, α ehris the importance weight of electronic health records, E ehr is the feature representation of electronic health records, α img is the importance weight of medical images, E img is the feature representation of medical images, α gen is the importance weight of genomic data, E gen Feature representation for genomics data;

[0037] Graph neural network (GNN) is introduced to process heterogeneous data sets and build a unified data representation space. The expression is:

[0038]

[0039] Among them, H (l+1) is the node feature matrix after GNN update, is the inverse square root of the degree matrix, is the adjacency matrix after adding self-loops, H (l) is the node feature matrix of the current layer, W (l) is the learnable weight matrix;

[0040] After the above steps, an optimized dataset containing multiple information is obtained, that is, a dataset after multimodal fusion processing.

[0041] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, wherein: based on the data set processed by multimodal fusion, the diagnostic model is optimized using cross-validation technology, and the data set processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic result, the specific steps are as follows:

[0042] Set accuracy as the evaluation indicator for evaluating model performance;

[0043] The K-fold cross-validation technique is used to optimize the diagnostic model. During each cross-validation process, the average accuracy of the model is calculated, and the expression is:

[0044]

[0045] in, is the average accuracy, K is the number of subsets, M i is the accuracy score in the i-th cross validation;

[0046] Introducing grid search and random search techniques to explore different hyperparameter combinations and find the optimal configuration;

[0047] Suppose there are three hyperparameters that need to be adjusted: learning rate, batch size, and regularization coefficient;

[0048] For each hyperparameter combination, calculate the corresponding average accuracy and select the optimal hyperparameter combination. The expression is:

[0049]

[0050] Among them, η * is the optimal learning rate, B * is the optimal batch size, λ * is the optimal regularization coefficient, is the average accuracy under a specific hyperparameter combination;

[0051] Based on the optimized hyperparameter settings, the diagnostic model is retrained, and the dataset processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic results;

[0052] The optimized diagnosis result includes the final classification label and the confidence score of each category.

[0053] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, the steps of generating a diagnosis report based on the optimized diagnosis results using an explanatory algorithm are as follows:

[0054] Based on the optimized diagnostic results, the model predictions were explained using the interpretative algorithm SHAP;

[0055] For each input sample, calculate its corresponding SHAP value, the expression is:

[0056]

[0057] Where N is the set of all features, S is the feature subset, f(S) represents the model prediction value containing only the feature subset S, φ i (x) represents the contribution of the i-th feature to the model prediction results;

[0058] Generate a detailed diagnostic interpretation report based on the SHAP value;

[0059] The diagnostic interpretation report includes the final diagnostic result and the specific impact of each feature on the diagnostic result;

[0060] In the diagnostic interpretation report, the most important features and their corresponding SHAP values ​​are listed in order of their importance. The expression is:

[0061]

[0062] Among them, Report is the diagnosis interpretation report, is the i-th feature, φ i is the SHAP value of the i-th feature.

[0063] As a preferred embodiment of the computer-aided medical diagnosis method of the present invention, wherein: the key information in the diagnosis report is displayed using visualization technology to obtain a visual diagnosis result, the specific steps are:

[0064] Design an interactive user interface that allows doctors to view detailed diagnostic processes and the impact of key features;

[0065] Heat map: shows the contribution distribution of different features to model predictions;

[0066] 3D reconstructed images: For medical imaging data, 3D reconstruction technology is used to highlight the lesion area;

[0067] Time series analysis chart: For data that changes over time, draw a time series chart to show the changing trends of key indicators;

[0068] Integrate all visual elements to form a complete visual diagnostic result.

[0069] In a second aspect, the present invention provides a computer-aided medical diagnosis system, comprising:

[0070] Data preprocessing module, model building and training module, multimodal data fusion module, model optimization module and result interpretation and visualization module;

[0071] The data preprocessing module is used to collect raw data from different sources and perform cleaning, denoising and standardization on it to obtain a high-quality data set;

[0072] The model building and training module is used to build a diagnostic model based on high-quality data sets and deep learning algorithms, input high-quality data sets into the diagnostic model for training, and output preliminary diagnostic results;

[0073] The multimodal data fusion module is used to integrate the data used for the diagnosis results using a multimodal data fusion method, integrating data from different sources into a unified framework for analysis to obtain a data set after multimodal fusion processing;

[0074] The model optimization module is used to optimize the diagnostic model using cross-validation technology based on the data set after multimodal fusion processing;

[0075] The result interpretation and visualization module is used to generate a detailed diagnostic report based on the optimized diagnostic results using an explanatory algorithm, and to display key information using visualization technology to form a complete visual diagnostic result.

[0076] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the computer-assisted medical diagnosis method as described in the first aspect of the present invention is implemented.

[0077] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the computer-assisted medical diagnosis method as described in the first aspect of the present invention.

[0078] The beneficial effects of the present invention are: by constructing a diagnostic model based on high-quality data sets and deep learning algorithms, and inputting high-quality data sets into the diagnostic model, and outputting preliminary diagnostic results, it can not only automatically extract complex feature representations, but also prevent overfitting through regularization technology, thereby improving the generalization ability of the model. A multimodal data fusion method is used to integrate the data used for the diagnostic results, and data from different sources are integrated into a unified framework for analysis to obtain a data set after multimodal fusion processing. The multimodal data fusion strategy can make full use of the advantages of data from different sources and provide more comprehensive information support. The model performance is evaluated by the K-fold cross-validation technology, and the grid search and random search technologies are used to explore the best hyperparameter combination, and the optimized diagnostic model is retrained. The method not only takes into account the performance of the model on the training set, but also evaluates its generalization ability through the validation set, thereby ensuring the efficiency and reliability of the model in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of 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.

[0080] Figure 1 This is a flow chart of the computer-aided medical diagnosis method in Example 1.

[0081] Figure 2 Schematic diagram of the computer-aided medical diagnosis system in Example 1. DETAILED DESCRIPTION

[0082] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0083] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0084] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0085] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a computer-aided medical diagnosis method, comprising the following steps:

[0086] S1. Data collection and preprocessing methods are used to clean, denoise, and standardize patient medical data from different sources to obtain high-quality datasets.

[0087] Going a step further, raw data is collected from electronic health records, medical imaging, and genomic data;

[0088] Use natural language processing (NLP) to segment text data, remove stop words, and perform stemming operations. Calculate the importance score of each word using the TF-IDF algorithm to achieve data cleaning.

[0089] Apply Gaussian filtering to the image data to smooth the image and reduce noise;

[0090] The Z-Score normalization method was used to convert genomic data into a standard normal distribution and obtain a high-quality dataset;

[0091] It should be noted that in the data collection and preprocessing stage, in addition to cleaning, denoising and standardizing the original data, special emphasis is placed on using specialized processing methods for different types of data. For example, the use of the TF-IDF algorithm can not only effectively extract key information from text data, but also significantly reduce data dimensions and improve the efficiency of subsequent model training; the application of Gaussian filters helps to eliminate noise in medical images and improve image quality. The data preprocessing measures lay a solid foundation for subsequent steps and ensure the generation of high-quality data sets.

[0092] S2. Build a diagnostic model based on high-quality data sets and deep learning algorithms, input the high-quality data sets into the diagnostic model, and output the diagnostic results;

[0093] Furthermore, a diagnostic model is constructed based on high-quality datasets and convolutional neural networks (CNNs), which can be expressed as:

[0094]

[0095] Among them, L represents the total loss, y i and Represent the true label and predicted value of the i-th sample respectively, N is the total number of samples, λ is the regularization coefficient used to control the complexity of the model, Ω represents the weight space, Indicates the absolute value of the weight change rate;

[0096] Input high-quality data sets into the diagnostic model for training, the expression is:

[0097]

[0098] Where BN(x) represents the normalized output, x is the input data, E[x] and Var[x] are the mean and variance of the input data, respectively. It is a small constant for numerical stability, and γ and β are learnable parameters used to scale and translate the normalized data.

[0099] Input a high-quality data set into the trained diagnostic model and output the diagnostic result. The expression is:

[0100]

[0101] Among them, P(y|x) represents the probability that the model predicts the output as category y given the input data x, x is the input data, x i represents the i-th eigenvalue of the input data, f(x i ) is a nonlinear activation function, w i is the weight parameter associated with the i-th feature, σ is the Sigmoid function;

[0102] The preliminary diagnostic results obtained will serve as the basis for subsequent multimodal data fusion to further improve the comprehensiveness and accuracy of the diagnosis;

[0103] It should be noted that when building a diagnostic model based on a high-quality dataset, the deep learning architecture of convolutional neural network (CNN) was adopted, and the model parameters were optimized by defining the loss function and batch normalization formula. This not only improved the performance of the model on complex tasks, but also effectively prevented the occurrence of overfitting. In addition, the output was limited to the range of [0,1] through the Sigmoid function, making the model prediction results probabilistically interpretable, making it easier for doctors to understand and apply the preliminary diagnostic results, and providing a reliable basis for further multimodal data fusion.

[0104] S3. Use multimodal data fusion methods to integrate the data used for diagnostic results, integrate data from different sources into a unified framework for analysis, and obtain a multimodal fusion data set;

[0105] Furthermore, a multimodal fusion function is defined to integrate data from different sources;

[0106] Set up a fusion model based on the attention mechanism to dynamically adjust the importance weights of different modal data;

[0107] The importance weight of each modality is calculated using the attention mechanism, and the expression is:

[0108]

[0109] Among them, α i is the importance weight of the i-th mode, e i is the scoring function output of the i-th mode;

[0110] F=α ehr ·E ehr +α img ·E img +α gen ·E gen ;

[0111] Among them, F is the final fusion feature vector, α ehr is the importance weight of electronic health records, E ehr is the feature representation of electronic health records, α img is the importance weight of medical images, E img is the feature representation of medical images, α gen is the importance weight of genomic data, E gen Feature representation for genomics data;

[0112] Graph neural network (GNN) is introduced to process heterogeneous data sets and build a unified data representation space. The expression is:

[0113]

[0114] Among them, H (l+1) is the node feature matrix after GNN update, is the inverse square root of the degree matrix, is the adjacency matrix after adding self-loops, H (l) is the node feature matrix of the current layer, W (l) is the learnable weight matrix;

[0115] After the above steps, an optimized dataset containing multiple information is obtained, that is, a dataset after multimodal fusion processing;

[0116] It should be noted that in the process of multimodal data fusion, the attention mechanism is used to dynamically adjust the importance weights of data of different modalities, ensuring that each type of data plays an appropriate role in the final decision. In particular, the introduction of graph neural networks (GNNs) to process heterogeneous data sets can effectively capture the complex relationships between data and construct a unified data representation space. This not only improves the model's ability to understand multi-source data, but also enhances the comprehensiveness and accuracy of diagnostic results, and helps to discover potential disease patterns that are difficult to identify with a single modality.

[0117] S4. Based on the data set processed by multimodal fusion, the diagnostic model is optimized using cross-validation technology, and the data set processed by multimodal fusion is input into the optimized diagnostic model to obtain an optimized diagnostic result;

[0118] Furthermore, accuracy is set as the evaluation index for evaluating model performance;

[0119] The K-fold cross-validation technique is used to optimize the diagnostic model. During each cross-validation process, the average accuracy of the model is calculated, and the expression is:

[0120]

[0121] in, is the average accuracy, K is the number of subsets, M i is the accuracy score in the i-th cross validation;

[0122] Introducing grid search and random search techniques to explore different hyperparameter combinations and find the optimal configuration;

[0123] Suppose there are three hyperparameters that need to be adjusted: learning rate, batch size, and regularization coefficient;

[0124] For each hyperparameter combination, calculate the corresponding average accuracy and select the optimal hyperparameter combination. The expression is:

[0125]

[0126] Among them, η * is the optimal learning rate, B * is the optimal batch size, λ * is the optimal regularization coefficient, is the average accuracy under a specific hyperparameter combination;

[0127] Based on the optimized hyperparameter settings, the diagnostic model is retrained, and the dataset processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic results;

[0128] The optimized diagnosis result includes the final classification label and the confidence score of each category;

[0129] It should be noted that during the model optimization stage, the model performance was evaluated through K-fold cross-validation technology, and the optimal hyperparameter combination was explored using grid search and random search to find the model configuration that best suits the current task. The systematic optimization method not only improves the generalization ability and stability of the model, but also reduces the consumption of computing resources and improves operational efficiency. The optimized diagnostic model can handle new patients or rare cases more accurately, further enhancing the practicality and reliability of the system.

[0130] S5. Based on the optimized diagnostic results, a diagnostic report is generated using an explanatory algorithm, and key information in the diagnostic report is displayed using visualization technology to obtain a visual diagnostic result.

[0131] Furthermore, based on the optimized diagnostic results, the model predictions were explained using the interpretative algorithm SHAP;

[0132] For each input sample, calculate its corresponding SHAP value, the expression is:

[0133]

[0134] Where N is the set of all features, S is the feature subset, f(S) represents the model prediction value containing only the feature subset S, φ i (x) represents the contribution of the i-th feature to the model prediction results;

[0135] Generate a detailed diagnostic interpretation report based on the SHAP value;

[0136] The diagnostic interpretation report includes the final diagnostic result and the specific impact of each feature on the diagnostic result;

[0137] In the diagnostic interpretation report, the most important features and their corresponding SHAP values ​​are listed in order of their importance. The expression is:

[0138]

[0139] Among them, Report is the diagnosis interpretation report, is the i-th feature, φ i is the SHAP value of the i-th feature;

[0140] Use visualization technology to display key information in the diagnosis report and obtain visual diagnosis results;

[0141] Design an interactive user interface that allows doctors to view detailed diagnostic processes and the impact of key features;

[0142] Heat map: shows the contribution distribution of different features to model predictions;

[0143] 3D reconstructed images: For medical imaging data, 3D reconstruction technology is used to highlight the lesion area;

[0144] Time series analysis chart: For data that changes over time, draw a time series chart to show the changing trends of key indicators;

[0145] Integrate all visual elements to form a complete visual diagnostic result;

[0146] It should be noted that the explanatory algorithm SHAP is used to generate detailed diagnostic reports, and a variety of visualization techniques are used to display key information, so that doctors can intuitively understand the specific impact of each feature on the diagnostic results. For example, the heat map clearly shows the contribution distribution of different features to the model prediction, the three-dimensional reconstructed image highlights the diseased area in the medical image, and the time series analysis chart shows the trend of key indicators over time. The visualization tool not only enhances doctors' trust in the AI ​​system, but also promotes the development of personalized precision medicine and helps to formulate more scientific and reasonable treatment plans.

[0147] This embodiment also provides a computer-aided medical diagnosis system, comprising:

[0148] Data preprocessing module, model building and training module, multimodal data fusion module, model optimization module and result interpretation and visualization module;

[0149] The data preprocessing module is used to collect raw data from different sources and perform cleaning, denoising and standardization on it to obtain a high-quality data set;

[0150] The model building and training module is used to build a diagnostic model based on high-quality data sets and deep learning algorithms, input high-quality data sets into the diagnostic model for training, and output preliminary diagnostic results;

[0151] The multimodal data fusion module is used to integrate the data used for the diagnosis results using a multimodal data fusion method, integrating data from different sources into a unified framework for analysis to obtain a data set after multimodal fusion processing;

[0152] The model optimization module is used to optimize the diagnostic model using cross-validation technology based on the data set after multimodal fusion processing;

[0153] The result interpretation and visualization module is used to generate a detailed diagnostic report based on the optimized diagnostic results using an explanatory algorithm, and to display key information using visualization technology to form a complete visual diagnostic result.

[0154] This embodiment also provides a computer device suitable for computer-assisted medical diagnosis methods, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the computer-assisted medical diagnosis method proposed in the above embodiment.

[0155] The computer device may be a terminal, comprising a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner may be achieved through Wi-Fi, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a button, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse.

[0156] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the computer-assisted medical diagnosis method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0157] In summary, the present invention constructs a diagnostic model based on high-quality data sets and deep learning algorithms, inputs high-quality data sets into the diagnostic model, and outputs preliminary diagnostic results. It can not only automatically extract complex feature representations, but also prevent overfitting through regularization technology, thereby improving the generalization ability of the model. A multimodal data fusion method is used to integrate the data used for the diagnostic results, and data from different sources are integrated into a unified framework for analysis to obtain a data set after multimodal fusion processing. The multimodal data fusion strategy can make full use of the advantages of data from different sources and provide more comprehensive information support. The model performance is evaluated by K-fold cross-validation technology, and the optimal hyperparameter combination is explored using grid search and random search techniques. The optimized diagnostic model is retrained. The method not only takes into account the performance of the model on the training set, but also evaluates its generalization ability through the validation set, thereby ensuring the efficiency and reliability of the model in practical applications.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A computer-aided medical diagnosis method, characterized in that: include: Data collection and preprocessing methods are used to clean, denoise, and standardize patient medical data from different sources to obtain high-quality datasets; Build a diagnostic model based on high-quality data sets and deep learning algorithms, input high-quality data sets into the diagnostic model, and output diagnostic results; The multimodal data fusion method is used to integrate the data used for the diagnosis results, integrating data from different sources into a unified framework for analysis to obtain a multimodal fusion processed data set; Based on the data set processed by multimodal fusion, the cross-validation technology is used to optimize the diagnostic model, and the data set processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic results; Based on the optimized diagnostic results, an explanatory algorithm is used to generate a diagnostic report, and visualization technology is used to display the key information in the diagnostic report to obtain a visual diagnostic result.

2. The computer-aided medical diagnosis method according to claim 1, wherein: The data collection and preprocessing method is used to clean, denoise, and standardize patient medical data from different sources to obtain a high-quality data set. The specific steps are as follows: Collect raw data from electronic health records, medical imaging, and genomic data; Use natural language processing (NLP) to segment text data, remove stop words, and perform stemming operations. Calculate the importance score of each word using the TF-IDF algorithm to achieve data cleaning. Apply Gaussian filtering to the image data to smooth the image and reduce noise; The Z-Score normalization method was used to transform genomic data into a standard normal distribution and obtain a high-quality dataset.

3. The computer-aided medical diagnosis method according to claim 2, wherein: The diagnostic model is constructed based on high-quality data sets and deep learning algorithms, and the high-quality data sets are input into the diagnostic model to output the diagnostic results. The specific steps are as follows: A diagnostic model is constructed based on high-quality data sets and convolutional neural networks (CNNs), which can be expressed as: Among them, L represents the total loss, y i and Represent the true label and predicted value of the i-th sample respectively, N is the total number of samples, λ is the regularization coefficient used to control the complexity of the model, Ω represents the weight space, Indicates the absolute value of the weight change rate; Input high-quality data sets into the diagnostic model for training, the expression is: Where BN(x) represents the normalized output, x is the input data, E[x] and Var[x] are the mean and variance of the input data, respectively. It is a small constant for numerical stability, and γ and β are learnable parameters used to scale and translate the normalized data. Input a high-quality data set into the trained diagnostic model and output the diagnostic result. The expression is: Among them, P(y|x) represents the probability that the model predicts the output as category y given the input data x, x is the input data, x i represents the i-th eigenvalue of the input data, f(x i ) is a nonlinear activation function, w i is the weight parameter associated with the i-th feature, σ is the Sigmoid function; The preliminary diagnostic results obtained will serve as the basis for subsequent multimodal data fusion to further improve the comprehensiveness and accuracy of the diagnosis.

4. The computer-aided medical diagnosis method according to claim 3, wherein: The multimodal data fusion method is used to integrate the data used for the diagnosis results, and the data from different sources are integrated into a unified framework for analysis to obtain a data set after multimodal fusion processing. The specific steps are as follows: Define a multimodal fusion function to integrate data from different sources; Set up a fusion model based on the attention mechanism to dynamically adjust the importance weights of different modal data; The importance weight of each modality is calculated using the attention mechanism, and the expression is: Among them, α i is the importance weight of the i-th mode, e i is the scoring function output of the i-th mode; F=a ehr ·E ehr +a img ·E img +a gen ·E gen ; Among them, F is the final fusion feature vector, α ehr is the importance weight of electronic health records, E ehr is the feature representation of electronic health records, α img is the importance weight of medical images, E img is the feature representation of medical images, α gen is the importance weight of genomic data, E gen Feature representation for genomics data; Graph neural network (GNN) is introduced to process heterogeneous data sets and build a unified data representation space. The expression is: Among them, H (l+1) is the node feature matrix after GNN update, is the inverse square root of the degree matrix, is the adjacency matrix after adding self-loops, H (l) is the node feature matrix of the current layer, W (l) is the learnable weight matrix; After the above steps, an optimized dataset containing multiple information is obtained, that is, a dataset after multimodal fusion processing.

5. The computer-aided medical diagnosis method according to claim 4, wherein: The cross-validation technique is used to optimize the diagnostic model based on the data set processed by multimodal fusion, and the data set processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic result. The specific steps are as follows: Set accuracy as the evaluation indicator for evaluating model performance; The K-fold cross-validation technique is used to optimize the diagnostic model. During each cross-validation process, the average accuracy of the model is calculated, and the expression is: in, is the average accuracy, K is the number of subsets, M i is the accuracy score in the i-th cross validation; Introducing grid search and random search techniques to explore different hyperparameter combinations and find the optimal configuration; Suppose there are three hyperparameters that need to be adjusted: learning rate, batch size, and regularization coefficient; For each hyperparameter combination, calculate the corresponding average accuracy and select the optimal hyperparameter combination. The expression is: Among them, η * is the optimal learning rate, B * is the optimal batch size, λ * is the optimal regularization coefficient, is the average accuracy under a specific hyperparameter combination; Based on the optimized hyperparameter settings, the diagnostic model is retrained, and the dataset processed by multimodal fusion is input into the optimized diagnostic model to obtain the optimized diagnostic results; The optimized diagnosis result includes the final classification label and the confidence score of each category.

6. The computer-aided medical diagnosis method according to claim 5, wherein: The diagnostic report is generated based on the optimized diagnostic results using an explanatory algorithm, and the specific steps are as follows: Based on the optimized diagnostic results, the model predictions were explained using the interpretative algorithm SHAP; For each input sample, calculate its corresponding SHAP value, the expression is: Where N is the set of all features, S is the feature subset, f(S) represents the model prediction value containing only the feature subset S, φ i (x) represents the contribution of the i-th feature to the model prediction results; Generate a detailed diagnostic interpretation report based on the SHAP value; The diagnostic interpretation report includes the final diagnostic result and the specific impact of each feature on the diagnostic result; In the diagnostic interpretation report, the most important features and their corresponding SHAP values ​​are listed in order of their importance. The expression is: Among them, Report is the diagnosis interpretation report, is the i-th feature, φ i is the SHAP value of the i-th feature.

7. The computer-aided medical diagnosis method according to claim 6, wherein: The visualization technology is used to display the key information in the diagnosis report to obtain the visual diagnosis results. The specific steps are as follows: Design an interactive user interface that allows doctors to view detailed diagnostic processes and the impact of key features; Heat map: shows the contribution distribution of different features to model predictions; 3D reconstructed images: For medical imaging data, 3D reconstruction technology is used to highlight the lesion area; Time series analysis chart: For data that changes over time, draw a time series chart to show the changing trends of key indicators; Integrate all visual elements to form a complete visual diagnostic result.

8. A computer-aided medical diagnosis system based on the computer-aided medical diagnosis method according to any one of claims 1 to 7, characterized in that: include: Data preprocessing module, model building and training module, multimodal data fusion module, model optimization module and result interpretation and visualization module; The data preprocessing module is used to collect raw data from different sources and perform cleaning, denoising and standardization on it to obtain a high-quality data set; The model building and training module is used to build a diagnostic model based on high-quality data sets and deep learning algorithms, input high-quality data sets into the diagnostic model for training, and output preliminary diagnostic results; The multimodal data fusion module is used to integrate the data used for the diagnosis results using a multimodal data fusion method, integrating data from different sources into a unified framework for analysis to obtain a data set after multimodal fusion processing; The model optimization module is used to optimize the diagnostic model using cross-validation technology based on the data set after multimodal fusion processing; The result interpretation and visualization module is used to generate a detailed diagnostic report based on the optimized diagnostic results using an explanatory algorithm, and to display key information using visualization technology to form a complete visual diagnostic result.

Images