Computer-aided intracranial aneurysm information processing system and method

Abstract

The invention discloses an intracranial aneurysm information processing system and method based on computer assistance, and relates to the technical field of medical information processing. An intracranial aneurysm information processing system based on computer assistance comprises a basic intracranial aneurysm information processing module and a special intracranial aneurysm information processing module. According to the method, the intracranial aneurysm image data, the auxiliary information data and the basic information analysis model are combined, and the feature enhancement and fusion technology is utilized, so that the diagnosis accuracy of the intracranial aneurysm can be remarkably improved; through an analysis result, high-confidence diagnosis information can be obtained; under the condition of low confidence, the special information analysis model can further optimize the diagnosis process, ensure the accuracy of the diagnosis result and reduce errors caused by feature complexity or data noise through a secondary feature processing and special feature recognition technology.

Description

Technical Field

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

[0002] At present, the information processing methods of intracranial aneurysms mainly rely on traditional imaging methods, such as CT, MRI and angiography. Although these technologies can effectively detect the presence of aneurysms, they still have limitations in early detection and identification of intracranial aneurysms; traditional intracranial aneurysm information processing methods rely on the doctor's experience and judgment, and lack sufficient targeted training and data support for the identification of rare cranial aneurysms, resulting in low accuracy in prediction and diagnosis; in addition, there are fewer rare aneurysm samples, which affects the effectiveness and generalizability of the data model, further limiting the early prediction and personalized treatment of the disease.

[0003] Therefore, further improving diagnostic accuracy, processing key image features through feature processing and feature recognition technology, and optimizing the diagnostic decision-making process have become hot topics in current research. Summary of the invention

[0004] The present invention aims to provide a computer-aided intracranial aneurysm information processing system and method to improve the diagnostic accuracy.

[0005] The computer-aided intracranial aneurysm information processing method comprises the following steps: Obtain the patient's intracranial aneurysm imaging data and intracranial aneurysm auxiliary information data; perform analysis based on the intracranial aneurysm imaging data, intracranial aneurysm auxiliary information data and intracranial aneurysm basic information analysis model to obtain intracranial aneurysm diagnosis information results and intracranial aneurysm diagnosis information confidence; When the intracranial aneurysm diagnostic information confidence level is lower than the preset information confidence level threshold, the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data are input into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnostic information result; otherwise, no operation is performed; The intracranial aneurysm special information analysis model further screens and identifies key imaging features through secondary feature processing and special feature recognition technology; Perform follow-up operations on the patient based on the intracranial aneurysm diagnostic information results and / or the intracranial aneurysm special diagnostic information results.

[0006] As a preferred technical solution of the present invention, the basic information analysis model of intracranial aneurysm includes a data feature extraction layer, a feature processing layer and a feature recognition layer; The data feature extraction layer is used to convert the intracranial aneurysm auxiliary information data into intracranial aneurysm auxiliary embedding vector feature L, L = [L1 , L 2 , …, L n , …, L N ], L n represents the embedded vector feature of the nth auxiliary information data item in the auxiliary information data of intracranial aneurysm; the intracranial aneurysm image data is cut to obtain M intracranial aneurysm image data blocks P m , m=1,2,…,M; The feature processing layer is used to embed the intracranial aneurysm auxiliary vector feature L and all intracranial aneurysm image data blocks P m Perform feature processing to obtain information fusion features of intracranial aneurysms; The feature recognition layer is used to perform feature recognition based on the intracranial aneurysm information fusion features to obtain the intracranial aneurysm diagnosis information results and the intracranial aneurysm diagnosis information confidence.

[0007] As a preferred technical solution of the present invention, the specific steps of performing feature processing in the feature processing layer include: The feature processing layer includes a feature enhancement unit and a feature fusion unit; In the feature enhancement unit, using formula X mn =(L n W n )(P m W m ) T Calculate the intracranial aneurysm feature correlation matrix X mn ; Among them, W n is the first weight matrix for pre-training intracranial aneurysms, W m To pre-train the second weight matrix for intracranial aneurysms; Using the formula Calculate the distribution of attention associated with intracranial aneurysm features ; Indicates the encoding vector corresponding to the current position; Indicates tuning of hyperparameters; is the transpose of the intracranial aneurysm feature association matrix; Using the formula For intracranial aneurysm image data block P m Perform feature enhancement operation to obtain enhanced intracranial aneurysm image feature P m ';P m '; Represents Hadamard operation; In the feature fusion unit, all enhanced intracranial aneurysm image features P m ', and obtain the fusion features of intracranial aneurysm information.

[0008] As a preferred technical solution of the present invention, the specific steps of calculating the confidence of intracranial aneurysm diagnosis information in the feature recognition layer include: In the feature recognition layer, the pre-trained intracranial aneurysm recognition confidence function is used to predict the probability distribution of the output intracranial aneurysm diagnosis information results to obtain the intracranial aneurysm diagnosis information confidence; The pre-trained intracranial aneurysm identification confidence function is trained by the softmax function as the original function; The specific steps for training the feature recognition layer include: Collecting several groups of intracranial aneurysm information feature recognition training samples; each group of intracranial aneurysm information feature recognition training samples contains historically diagnosed intracranial aneurysm information fusion features and corresponding verified intracranial aneurysm diagnosis information results; combining several groups of intracranial aneurysm information feature recognition training samples to obtain an intracranial aneurysm information feature recognition training set; The intracranial aneurysm information feature recognition training set is input into the CNN model for model training to obtain the initial feature recognition layer; the initial feature recognition layer is evaluated, and if the initial feature recognition layer passes the model evaluation, the initial feature recognition layer is used as the feature recognition layer in the intracranial aneurysm basic information analysis model; otherwise, the intracranial aneurysm information feature recognition training set is used to continue model training.

[0009] As a preferred technical solution of the present invention, the intracranial aneurysm special information analysis model is constructed based on the intracranial aneurysm basic information analysis model, wherein the data feature extraction layer and the feature processing layer are retained; Add a secondary feature processing layer and a special feature recognition layer to the intracranial aneurysm special information analysis model; The secondary feature processing layer is used to enhance the intracranial aneurysm image feature P in the special feature recognition layer. m 'Perform secondary feature processing to obtain the imaging features for screening intracranial aneurysms T i , i=1, 2, …, I; I is the number of enhanced intracranial aneurysm imaging features retained after screening; The special feature recognition layer is used to screen the imaging features of intracranial aneurysms. i Perform feature recognition to obtain special diagnostic information results of intracranial aneurysms.

[0010] As a preferred technical solution of the present invention, the specific steps of performing secondary feature processing in the secondary feature processing layer include: Using formula S m =score(P m ') on the enhanced imaging features of intracranial aneurysms m 'Screen the suspiciousness and obtain the enhanced intracranial aneurysm imaging feature suspicious score S m; Among them, score (P m ')∈(0,1); When S m >θ, the corresponding enhanced intracranial aneurysm image feature P m 'Retained for screening imaging features of intracranial aneurysms T i Otherwise, the enhanced intracranial aneurysm imaging features P m 'To eliminate; where θ is the suspiciousness scoring threshold, score() is the pre-trained suspiciousness scoring function.

[0011] As a preferred technical solution of the present invention, the specific steps of training the special feature recognition layer include: Collect existing special intracranial aneurysm information feature recognition training samples; each special intracranial aneurysm information feature recognition training sample contains historically diagnosed special intracranial aneurysm information fusion features and corresponding verified special intracranial aneurysm diagnosis information results; For several original groups of intracranial aneurysm information feature recognition training samples, the improved GAN model is used to enhance the synthetic samples, and several groups of generated intracranial aneurysm information feature recognition training samples are obtained; Combine several groups of generated intracranial aneurysm information feature recognition training samples with existing special intracranial aneurysm information feature recognition training samples to obtain a special intracranial aneurysm information feature recognition training set, and use a self-supervised learning method to perform model training based on the special intracranial aneurysm information feature recognition training set to obtain a special feature recognition layer; The specific steps of using the improved GAN model to enhance synthetic samples include: In the GAN model, construct K resolution layers B k , k=1,2,…,K; each resolution layer B k Use independent sub-generators for image generation; At resolution layer B k In, k <K 0 When , no sample generation condition information is added; At resolution layer B k In, k>K 0 When adding sample generation condition information, the sample generation condition information is set based on the special intracranial aneurysm information feature recognition training sample; K 0 is the preset parameter; The intracranial aneurysm information feature recognition training samples are input into the improved GAN model for synthetic sample enhancement to obtain several groups of generated intracranial aneurysm information feature recognition training samples.

[0012] Computer-aided intracranial aneurysm information processing system includes: The basic intracranial aneurysm information processing module includes a data acquisition unit and an information identification unit; the data acquisition unit is used to acquire the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data of the patient; the information identification unit is used to analyze the intracranial aneurysm image data, the intracranial aneurysm auxiliary information data and the intracranial aneurysm basic information analysis model to obtain the intracranial aneurysm diagnosis information results and the intracranial aneurysm diagnosis information confidence; The special intracranial aneurysm information processing module includes a special information identification unit; the special information identification unit is used to input the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnosis information results when the intracranial aneurysm diagnosis information confidence level is lower than the preset information confidence level threshold; otherwise, no operation is performed; the intracranial aneurysm special information analysis model further screens and identifies key image features through secondary feature processing and special feature recognition technology; and subsequent operations are performed on the patient based on the intracranial aneurysm diagnosis information results and / or the intracranial aneurysm special diagnosis information results.

[0013] The present invention has the following advantages: The present invention combines intracranial aneurysm imaging data, auxiliary information data and basic information analysis models, and utilizes feature enhancement and fusion technology to significantly improve the diagnostic accuracy of intracranial aneurysms, help doctors identify and judge potential aneurysm cases earlier, and provide a more powerful reference basis; high-confidence diagnostic information can be obtained through the analysis results of the model; in low-confidence cases, the application of special information analysis models can further screen out key imaging features and reduce the possibility of missed diagnosis; special information analysis models can further optimize the diagnostic process through secondary feature processing and special feature recognition technology, ensure the accuracy of diagnostic results, and reduce errors caused by feature complexity or data noise.

[0014] The present invention evaluates the correlation between auxiliary information and image data blocks by calculating the intracranial aneurysm feature association matrix in the feature enhancement unit. The calculation of this association matrix can help the model capture the complex relationship between the data, thereby providing more accurate feature enhancement for the image data; in the enhancement process, the image data is combined with the corresponding weight matrix through the Hadamard operation, which further improves the feature expression ability of the image data; by calculating the feature association attention distribution and applying the Softmax function, the model can allocate attention between different features. This operation effectively improves the model's ability to focus on important features when processing intracranial aneurysm data. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1This is a schematic diagram of the structure of a computer-aided intracranial aneurysm information processing system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] Embodiment 1, a computer-aided intracranial aneurysm information processing method, comprising the following steps: Obtain the patient's intracranial aneurysm imaging data and intracranial aneurysm auxiliary information data; perform analysis based on the intracranial aneurysm imaging data, intracranial aneurysm auxiliary information data and intracranial aneurysm basic information analysis model to obtain intracranial aneurysm diagnosis information results and intracranial aneurysm diagnosis information confidence; The basic information analysis model of intracranial aneurysm improves the diagnostic accuracy by combining auxiliary information data with image features and using feature enhancement and fusion technology, providing reference information for early diagnosis and treatment; The intracranial aneurysm auxiliary information data contains a variety of data related to the patient's condition. In addition to imaging data, it also involves the patient's clinical, genetic, hematological and other information; for example, age, gender and family history, clinical history such as hypertension, smoking history and arteriosclerosis, and biomarker data, such as blood lipid levels, inflammatory indicators and blood sugar levels; in addition, it also involves genetic data, including gene mutations and family genetic factors, imaging information such as CT / MRI scan reports and angiography data, and hemodynamic data, such as vascular wall stress analysis and vascular curvature. These auxiliary information data combined with imaging data help provide a more comprehensive analysis and support accurate intracranial aneurysm diagnosis and treatment decisions; intracranial aneurysm imaging data generally refers to image data about intracranial aneurysms obtained through medical imaging technology (such as CT scans, MRI scans, etc.), which are mainly used to help doctors identify and evaluate the existence, size, location and other important characteristics of aneurysms; The basic information analysis model of intracranial aneurysm includes data feature extraction layer, feature processing layer and feature recognition layer; The data feature extraction layer is used to convert the intracranial aneurysm auxiliary information data into intracranial aneurysm auxiliary embedding vector feature L, L = [L 1 , L 2 , …, L n , …, L N ], L n represents the embedded vector feature of the nth auxiliary information data item in the auxiliary information data of intracranial aneurysm; the intracranial aneurysm image data is cut to obtain M intracranial aneurysm image data blocks P m, m=1, 2, …, M; the main function of the data feature extraction layer is to convert the auxiliary information data of intracranial aneurysms into embedded vector features, and to perform cropping on the image data of intracranial aneurysms to ensure that the subsequent steps can effectively process different types of input data; the auxiliary information data is converted into embedded vector features through a specific processing method. This process usually uses neural networks or other embedding technologies to convert high-dimensional, structured auxiliary information (such as the patient's age, medical history, blood pressure, etc.) into low-dimensional vectors. These vectors can retain the key information of the original information while reducing redundancy and facilitating subsequent processing; the converted embedded vector features are recorded as the auxiliary embedded vector features L of intracranial aneurysms, where each component L n Represents the embedded features of the nth information item in the auxiliary information data of intracranial aneurysm; for the processing of intracranial aneurysm image data, the image data usually has a high spatial dimension, so it is necessary to use cropping technology to divide the image data into multiple smaller image blocks. These image blocks can be small-sized images or image areas, which are convenient for subsequent feature extraction. This process converts the image data into a set of multiple data blocks, so that each image block can be processed and analyzed separately; The feature processing layer is used to embed the intracranial aneurysm auxiliary vector feature L and all intracranial aneurysm image data blocks P m Perform feature processing to obtain the information fusion features of intracranial aneurysms; the main function of the feature processing layer is to further process and fuse the embedded vector features and image data blocks from the data feature extraction layer, and its goal is to effectively combine the auxiliary information and image data, so as to more comprehensively describe the various features of intracranial aneurysms; The feature recognition layer is used to perform feature recognition based on the fusion features of intracranial aneurysm information, and obtain the intracranial aneurysm diagnosis information results and intracranial aneurysm diagnosis information confidence; the feature recognition layer is the core part of the model, and its main purpose is to diagnose intracranial aneurysms based on the fusion features obtained in the feature processing layer. Through the analysis and recognition of the fusion features, the model can determine whether the patient has intracranial aneurysms and further evaluate its severity or diagnostic category; The data feature extraction layer can fully capture information of different dimensions by converting the auxiliary information data of intracranial aneurysms into embedded vector features and cutting and processing the image data. This multi-level and multi-angle feature extraction method provides strong data support for subsequent analysis. The feature processing layer fuses the auxiliary information embedded features and image data blocks, so that the model can simultaneously process structural information and auxiliary information, thereby comprehensively describing the characteristics of intracranial aneurysms. This information fusion can enhance the model's ability to identify and diagnose complex diseases. The feature recognition layer can help doctors more accurately assess the condition of intracranial aneurysms by making diagnoses based on fused features. The model can output clear diagnostic results and give confidence levels, providing doctors with reliable decision support and further improving the efficiency and accuracy of clinical diagnosis. The specific steps of feature processing in the feature processing layer include: The feature processing layer includes a feature enhancement unit and a feature fusion unit; In the feature enhancement unit, using formula X mn =(L n W n )(P m W m ) T Calculate the intracranial aneurysm feature correlation matrix X mn ; Among them, W n is the first weight matrix for pre-training intracranial aneurysms, W m To pre-train the second weight matrix for intracranial aneurysms; Using the formula Calculate the distribution of attention associated with intracranial aneurysm features ; Indicates the encoding vector corresponding to the current position; is the transpose of the intracranial aneurysm feature association matrix; Hyperparameter adjustment is performed by professional technicians based on actual conditions. Hyperparameter adjustment is usually based on some factors in the actual application scenario. For example, different data sets may have different noise levels or feature distributions. It can help the model better adapt to the characteristics of the data; if the diagnostic task is more complex, the model may need more adjustments to improve accuracy, so it needs to be carefully adjusted. To balance the sensitivity and robustness of the model; according to the requirements of diagnostic accuracy and computational efficiency, The value can affect the model's balance between processing speed and accuracy; Using the formula For intracranial aneurysm image data block P m Perform feature enhancement operation to obtain enhanced intracranial aneurysm image feature P m';in, Represents Hadamard operation; In the feature fusion unit, all enhanced intracranial aneurysm image features P m ', obtain the information fusion features of intracranial aneurysm; In the feature enhancement unit, the correlation between the auxiliary information and the image data block is first evaluated by calculating the intracranial aneurysm feature association matrix. The calculation of this association matrix can help the model capture the complex relationship between the data, and then provide more accurate feature enhancement for the image data; in the enhancement process, the image data is combined with the corresponding weight matrix through the Hadamard operation, which further improves the feature expression ability of the image data; by calculating the feature association attention distribution and applying the Softmax function, the model can allocate attention between different features. This operation effectively improves the model's ability to focus on important features when processing intracranial aneurysm data; by adjusting the hyperparameters , the model can flexibly control the importance of different features, so that the focus on key features is more accurate, and the feature processing effect is further optimized; the feature fusion unit fuses all enhanced image features to form intracranial aneurysm information fusion features. This step enables the model to obtain a more comprehensive and efficient feature representation by integrating multi-faceted enhanced information, thereby improving the model's comprehensive diagnostic ability for intracranial aneurysms; The process of training the first weight matrix of pre-trained intracranial aneurysm and the second weight matrix of pre-trained intracranial aneurysm involves using a large amount of labeled image data and auxiliary information to perform end-to-end training through a deep learning model; first, the weight matrix is ​​initialized, and these weights represent the potential association between the image data and the auxiliary information; then, the image data and auxiliary information of the intracranial aneurysm are input into the model, the loss function is calculated through the back propagation algorithm, and the weight matrix is ​​adjusted according to the actual diagnosis results; during the training process, the optimization algorithm gradually updates the weight matrix so that the model can more effectively learn the relationship between the image features and the auxiliary information, thereby improving the accuracy and efficiency of the diagnosis, and ultimately achieving efficient and accurate diagnostic capabilities; the entire process requires a large amount of data and iterative training to ensure that the weight matrix can effectively capture the complex characteristics and patterns of intracranial aneurysms; The specific steps of calculating the confidence of intracranial aneurysm diagnosis information in the feature recognition layer include: In the feature recognition layer, the pre-trained intracranial aneurysm recognition confidence function is used to predict the probability distribution of the output intracranial aneurysm diagnosis information results to obtain the intracranial aneurysm diagnosis information confidence; The pre-trained intracranial aneurysm identification confidence function is trained by the softmax function as the original function; The use of the pre-trained confidence function for intracranial aneurysm identification can be carried out in the following steps: collect several sets of training data, each set of data includes: feature information of intracranial aneurysms and corresponding verified diagnostic results, organize these data into training sets for training models; define a neural network model, which can be CNN (Convolutional Neural Networks) to extract features from image data, and use the softmax function in the last layer to convert the features of the network output into probability distributions for predicting the classification results of intracranial aneurysms; use the cross entropy loss function, which is suitable for classification problems, to calculate the difference between the predicted probability and the actual label; input the training sample, calculate the output obtained by the neural network, that is, the predicted category probability distribution, and calculate the loss value through the cross entropy loss function based on the model output and the actual label; calculate the gradient through the back propagation algorithm, and update the parameters in the network to minimize the loss function; when the training process is completed and passed the evaluation, the model will be able to output the diagnosis results of intracranial aneurysms and give the corresponding confidence; The specific steps for training the feature recognition layer include: Collecting several groups of intracranial aneurysm information feature recognition training samples; each group of intracranial aneurysm information feature recognition training samples contains historically diagnosed intracranial aneurysm information fusion features and corresponding verified intracranial aneurysm diagnosis information results; combining several groups of intracranial aneurysm information feature recognition training samples to obtain an intracranial aneurysm information feature recognition training set; The intracranial aneurysm information feature recognition training set is input into the CNN model for model training to obtain the initial feature recognition layer; the initial feature recognition layer is evaluated, and if the initial feature recognition layer passes the model evaluation, the initial feature recognition layer is used as the feature recognition layer in the intracranial aneurysm basic information analysis model; otherwise, the intracranial aneurysm information feature recognition training set is used to continue model training; By using the pre-trained recognition confidence function, the model can predict the probability distribution of the diagnosis results of intracranial aneurysms. This method can not only provide specific diagnostic results, but also give the confidence of each result, helping doctors understand the reliability of the model's predictions and enhancing the transparency of the diagnostic process. The training feature recognition layer continuously inputs historical diagnostic data and verified feature recognition samples to gradually optimize the model. This training process allows the model to adapt to different types of intracranial aneurysm data, improving the adaptability and accuracy of the model. The model can automatically adjust until it meets the optimal performance through the evaluation feedback mechanism, reducing the need for human intervention. By combining historical diagnostic data and feature fusion information, the model can better understand and extract the key features of intracranial aneurysms. During the training process, CNN is used for model training, so that the model can automatically learn the deep relationship between features, thereby greatly improving the ability to process complex image data. When the confidence of the intracranial aneurysm diagnostic information is lower than the preset information confidence threshold, the intracranial aneurysm imaging data and the intracranial aneurysm auxiliary information data are input into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnostic information results; otherwise, no operation is performed; the preset information confidence threshold is set by professional and technical personnel according to actual conditions. For example, the technician can set the threshold based on the distribution of historical data and the accuracy of the diagnostic results to ensure that further analysis is performed only when the confidence of the model is high enough; secondly, the setting of the threshold should also consider the fault tolerance requirements in actual applications. For example, in some cases, the threshold may need to be lowered to capture more potential aneurysm cases, while in other cases, a higher confidence may be required to avoid misdiagnosis; in addition, the threshold can be dynamically adjusted with the progress of model training and the feedback of actual diagnostic results to optimize the diagnostic performance; at the same time, ensure that the set threshold is determined based on the experience of professional and technical personnel and the requirements of specific medical scenarios; The special information analysis model for intracranial aneurysms further screens and identifies key imaging features through secondary feature processing and special feature recognition technology, and optimizes the diagnostic decision-making process through secondary feature processing and special feature recognition technology; The special information analysis model of intracranial aneurysm is constructed based on the basic information analysis model of intracranial aneurysm, wherein the data feature extraction layer and feature processing layer are retained; Add a secondary feature processing layer and a special feature recognition layer to the intracranial aneurysm special information analysis model; The secondary feature processing layer is used to enhance the intracranial aneurysm image feature P in the special feature recognition layer. m 'Perform secondary feature processing to obtain the imaging features for screening intracranial aneurysms T i , i=1, 2, …, I; I is the number of enhanced intracranial aneurysm imaging features retained after screening; The special feature recognition layer is used to screen the imaging features of intracranial aneurysms. i Perform feature recognition to obtain special diagnostic information results of intracranial aneurysms; By setting the information confidence threshold, the model can flexibly determine when further analysis is needed; if the confidence of the initial diagnosis is lower than the set threshold, the model will automatically input the data into the special information analysis model for more detailed screening and diagnosis. This mechanism ensures that additional analysis is only performed in uncertain cases, thereby improving the efficiency of the model and reducing unnecessary calculations; the secondary feature processing layer further screens the enhanced image features and optimizes the expression of information. This step enhances the model's ability to distinguish in complex situations by removing redundant features and retaining the most useful features; in this way, the model can more accurately identify key features when faced with complex or ambiguous cases; through the special feature recognition layer, the screened image features are further identified and diagnosed, making the final diagnosis more accurate. By strengthening the screening and optimization of features, especially when the image data is more complex, the quality and accuracy of diagnostic decisions can be further improved; through the technology of secondary feature processing and special feature recognition, even when the initial model diagnosis result is uncertain, the special information analysis model can perform a more in-depth analysis to further reduce the risk of misdiagnosis and missed diagnosis, which is crucial for early detection and accurate treatment of intracranial aneurysms; The specific steps of performing secondary feature processing in the secondary feature processing layer include: Using formula S m =score(P m ') on the enhanced imaging features of intracranial aneurysms m 'Screen the suspiciousness and obtain the enhanced intracranial aneurysm imaging feature suspicious score S m ; Among them, score (P m ')∈(0,1); When S m >θ, the corresponding enhanced intracranial aneurysm image feature P m 'Retained for screening imaging features of intracranial aneurysms T i Otherwise, the enhanced intracranial aneurysm imaging features P m 'To eliminate; where θ is the suspiciousness scoring threshold, score() is the pre-trained suspiciousness scoring function; By using the formula to screen the suspicion of enhanced intracranial aneurysm imaging features, the suspicion score of each imaging feature can be calculated; the output value of the scoring function is between (0, 1), which enables the model to make detailed distinctions between different features and give priority to those features that are more likely to be related to intracranial aneurysms, thereby reducing the risk of misdiagnosis; by setting a suspicion score threshold, the model can automatically screen out the most critical imaging features; when the score exceeds the threshold, the feature will be retained for subsequent analysis, otherwise it will be eliminated; this mechanism automatically removes inefficient or irrelevant features to ensure that subsequent analysis focuses on the features with the most diagnostic value, thereby improving the efficiency and accuracy of feature selection; secondary feature processing not only improves diagnostic efficiency, but also reduces redundancy and noise in the data by eliminating irrelevant or unreliable features, which helps reduce the risk of model overfitting and makes the final model more robust; The specific steps for training the special feature recognition layer include: Collect existing special intracranial aneurysm information feature recognition training samples; each special intracranial aneurysm information feature recognition training sample contains historically diagnosed special intracranial aneurysm information fusion features and corresponding verified special intracranial aneurysm diagnosis information results; The training samples for special intracranial aneurysm information feature recognition are rare types of intracranial aneurysm cases, such as aneurysms located in the brainstem or deep brain regions, or aneurysms combined with cerebral artery malformations. The number of special intracranial aneurysm information feature recognition training samples is small, and the number of samples needs to be increased. For several original sets of intracranial aneurysm information feature recognition training samples, the improved GAN (Generative Adversarial Network) model was used to enhance the synthetic samples, and several sets of generated intracranial aneurysm information feature recognition training samples were obtained; some rare types of aneurysms show atypical imaging features and may be difficult to detect or diagnose through conventional imaging methods; Combine several groups of generated intracranial aneurysm information feature recognition training samples with existing special intracranial aneurysm information feature recognition training samples to obtain a special intracranial aneurysm information feature recognition training set, and use a self-supervised learning method to perform model training based on the special intracranial aneurysm information feature recognition training set to obtain a special feature recognition layer; The specific steps of using the improved GAN model to enhance synthetic samples include: In the GAN model, construct K resolution layers B k , k=1,2,…,K; each resolution layer B k Use independent sub-generators for image generation; At resolution layer B k In, k <K 0When , no sample generation condition information is added; At resolution layer B k In, k>K 0 When adding sample generation condition information, the sample generation condition information is set based on the special intracranial aneurysm information feature recognition training sample; K 0 K is a preset parameter, which is set by professional technicians according to the actual situation. It is determined comprehensively based on the complexity of the data set, the specific requirements of the task, the constraints of computing resources, and the expected training effect. If the features involved in the task are more complex and different resolution layers have a significant impact on the generation of samples, then K 0 Can be set to a lower value, which can ensure that the earlier resolution layers begin to introduce conditional information, thereby helping the model to better learn and generate data with specific characteristics; if the training samples are scarce and more conditions are needed to guide the generation, then K 0 The value can be set relatively high to ensure that higher resolution generated layers can add additional conditional information to help the model accurately generate samples that meet specific requirements; The intracranial aneurysm information feature recognition training samples are input into the improved GAN model for synthetic sample enhancement to obtain several groups of intracranial aneurysm information feature recognition training samples; By using the improved GAN model to enhance synthetic samples, additional training data can be generated. This process increases the diversity of the training set. In particular, when the existing samples are limited, synthetic data can supplement the scarcity of data and improve the robustness of the model. By generating more diverse training samples, especially features learned from real data, this helps to improve the recognition ability of the special feature recognition layer for complex and subtle intracranial aneurysm features. In particular, by introducing conditional information at a higher resolution level, the model can learn and generate specific features more accurately, enhancing the model's sensitivity to key features. By training the model through a self-supervised learning method, the model can learn autonomously from the data without relying on a large amount of manually labeled data, which not only speeds up the training speed and reduces the labeling cost, but also makes the model more flexible to adapt to different diagnostic situations and new data patterns. By introducing multiple resolution layers in the GAN model, especially combining sample generation conditional information at low and high resolution levels, the model can learn features at different levels in a more detailed manner. This hierarchical generation method can extract valuable features at different resolutions and improve the ability to capture detailed features. Follow-up operations are performed on the patient based on the results of the intracranial aneurysm diagnostic information and / or the results of the special diagnostic information of intracranial aneurysm, such as further imaging examinations, angiography, regular monitoring of the patient, and development of personalized treatment plans based on the type and size of the aneurysm; if the diagnostic results indicate that the risk of aneurysm is high, surgical treatment such as endovascular treatment or craniotomy may be required, or in some cases aneurysm embolization or clipping may be selected for treatment; in addition, for low-risk or smaller aneurysms, the doctor may recommend regular review and monitoring to evaluate whether the aneurysm has changed, and for patients with a genetic predisposition, genetic counseling or preventive treatment may also be involved.

[0018] Example 2, computer-assisted intracranial aneurysm information processing system, see Figure 1 As shown, including: The basic intracranial aneurysm information processing module includes a data acquisition unit and an information identification unit; the data acquisition unit is used to acquire the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data of the patient; the information identification unit is used to analyze the intracranial aneurysm image data, the intracranial aneurysm auxiliary information data and the intracranial aneurysm basic information analysis model to obtain the intracranial aneurysm diagnosis information results and the intracranial aneurysm diagnosis information confidence; The special intracranial aneurysm information processing module includes a special information identification unit; the special information identification unit is used to input the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnosis information results when the intracranial aneurysm diagnosis information confidence level is lower than the preset information confidence level threshold; otherwise, no operation is performed; the intracranial aneurysm special information analysis model further screens and identifies key image features through secondary feature processing and special feature recognition technology; and subsequent operations are performed on the patient based on the intracranial aneurysm diagnosis information results and / or the intracranial aneurysm special diagnosis information results.

[0019] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A computer-aided intracranial aneurysm information processing method, characterized in that: The following steps are involved: Obtain the patient's intracranial aneurysm imaging data and intracranial aneurysm auxiliary information data; perform analysis based on the intracranial aneurysm imaging data, intracranial aneurysm auxiliary information data and intracranial aneurysm basic information analysis model to obtain intracranial aneurysm diagnosis information results and intracranial aneurysm diagnosis information confidence; When the intracranial aneurysm diagnostic information confidence level is lower than the preset information confidence level threshold, the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data are input into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnostic information result; otherwise, no operation is performed; The intracranial aneurysm special information analysis model further screens and identifies key imaging features through secondary feature processing and special feature recognition technology; Perform follow-up operations on the patient based on the intracranial aneurysm diagnostic information results and / or the intracranial aneurysm special diagnostic information results.

2. The computer-aided intracranial aneurysm information processing method according to claim 1, characterized in that: The basic information analysis model of intracranial aneurysm includes data feature extraction layer, feature processing layer and feature recognition layer; The data feature extraction layer is used to convert the intracranial aneurysm auxiliary information data into intracranial aneurysm auxiliary embedding vector feature L, L = [L1, L2, ..., L n , …, L N ], L n represents the embedded vector feature of the nth auxiliary information data item in the auxiliary information data of intracranial aneurysm; the intracranial aneurysm image data is cut to obtain M intracranial aneurysm image data blocks P m , m=1,2,…,M; The feature processing layer is used to embed the intracranial aneurysm auxiliary vector feature L and all intracranial aneurysm image data blocks P m Perform feature processing to obtain information fusion features of intracranial aneurysms; The feature recognition layer is used to perform feature recognition based on the intracranial aneurysm information fusion features to obtain the intracranial aneurysm diagnosis information results and the intracranial aneurysm diagnosis information confidence.

3. The computer-aided intracranial aneurysm information processing method according to claim 2, characterized in that: The specific steps of feature processing in the feature processing layer include: The feature processing layer includes a feature enhancement unit and a feature fusion unit; In the feature enhancement unit, using formula X mn =(L n W n )(P m W m ) T Calculate the intracranial aneurysm feature correlation matrix X mn ; Among them, W n is the first weight matrix for pre-training intracranial aneurysms, W m To pre-train the second weight matrix for intracranial aneurysms; Using the formula Calculate the distribution of attention associated with intracranial aneurysm features ; Indicates the encoding vector corresponding to the current position; Indicates tuning of hyperparameters; is the transpose of the intracranial aneurysm feature association matrix; Using the formula For intracranial aneurysm image data block P m Perform feature enhancement operation to obtain enhanced intracranial aneurysm image feature P m ';in, Represents Hadamard operation; In the feature fusion unit, all enhanced intracranial aneurysm image features P m ', and obtain the fusion features of intracranial aneurysm information.

4. The computer-aided intracranial aneurysm information processing method according to claim 3, characterized in that: The specific steps of calculating the confidence of intracranial aneurysm diagnosis information in the feature recognition layer include: In the feature recognition layer, the pre-trained intracranial aneurysm recognition confidence function is used to predict the probability distribution of the output intracranial aneurysm diagnosis information results to obtain the intracranial aneurysm diagnosis information confidence; The pre-trained intracranial aneurysm identification confidence function is trained by the softmax function as the original function; The specific steps for training the feature recognition layer include: Collecting several groups of intracranial aneurysm information feature recognition training samples; each group of intracranial aneurysm information feature recognition training samples contains historically diagnosed intracranial aneurysm information fusion features and corresponding verified intracranial aneurysm diagnosis information results; combining several groups of intracranial aneurysm information feature recognition training samples to obtain an intracranial aneurysm information feature recognition training set; The intracranial aneurysm information feature recognition training set is input into the CNN model for model training to obtain the initial feature recognition layer; the initial feature recognition layer is evaluated, and if the initial feature recognition layer passes the model evaluation, the initial feature recognition layer is used as the feature recognition layer in the intracranial aneurysm basic information analysis model; otherwise, the intracranial aneurysm information feature recognition training set is used to continue model training.

5. The computer-aided intracranial aneurysm information processing method according to claim 4, characterized in that: The special information analysis model of intracranial aneurysm is constructed based on the basic information analysis model of intracranial aneurysm, wherein the data feature extraction layer and feature processing layer are retained; Add a secondary feature processing layer and a special feature recognition layer to the intracranial aneurysm special information analysis model; The secondary feature processing layer is used to enhance the intracranial aneurysm image feature P in the special feature recognition layer. m 'Perform secondary feature processing to obtain the imaging features for screening intracranial aneurysms T i , i=1, 2, …, I; I is the number of enhanced intracranial aneurysm imaging features retained after screening; The special feature recognition layer is used to screen the imaging features of intracranial aneurysms. i Perform feature recognition to obtain special diagnostic information results of intracranial aneurysms.

6. The computer-aided intracranial aneurysm information processing method according to claim 5, characterized in that: The specific steps of performing secondary feature processing in the secondary feature processing layer include: Using formula S m =score(P m ') on the enhanced imaging features of intracranial aneurysms m 'Screen the suspiciousness and obtain the enhanced intracranial aneurysm imaging feature suspicious score S m ; Among them, score (P m ')∈(0,1); When S m >θ, the corresponding enhanced intracranial aneurysm image feature P m 'Retained for screening imaging features of intracranial aneurysms T i Otherwise, the enhanced intracranial aneurysm imaging features P m 'To eliminate; where θ is the suspiciousness scoring threshold, score() is the pre-trained suspiciousness scoring function.

7. The computer-aided intracranial aneurysm information processing method according to claim 6, characterized in that: The specific steps for training the special feature recognition layer include: Collect existing special intracranial aneurysm information feature recognition training samples; each special intracranial aneurysm information feature recognition training sample contains historically diagnosed special intracranial aneurysm information fusion features and corresponding verified special intracranial aneurysm diagnosis information results; For several original groups of intracranial aneurysm information feature recognition training samples, the improved GAN model is used to enhance the synthetic samples, and several groups of generated intracranial aneurysm information feature recognition training samples are obtained; Combine several groups of generated intracranial aneurysm information feature recognition training samples with existing special intracranial aneurysm information feature recognition training samples to obtain a special intracranial aneurysm information feature recognition training set, and use a self-supervised learning method to perform model training based on the special intracranial aneurysm information feature recognition training set to obtain a special feature recognition layer; The specific steps of using the improved GAN model to enhance synthetic samples include: In the GAN model, construct K resolution layers B k , k=1,2,…,K; each resolution layer B k Use independent sub-generators for image generation; In resolution layer B k when k < K0, no sample generation condition information is added; At resolution layer B k In the example, when k>K0, sample generation condition information is added; the sample generation condition information is set based on special intracranial aneurysm information feature recognition training samples; K0 is the preset parameter; The intracranial aneurysm information feature recognition training samples are input into the improved GAN model for synthetic sample enhancement to obtain several groups of generated intracranial aneurysm information feature recognition training samples.

8. A computer-aided intracranial aneurysm information processing system, characterized in that: The system applies the computer-aided intracranial aneurysm information processing method according to any one of claims 1 to 7, comprising: The basic intracranial aneurysm information processing module includes a data acquisition unit and an information identification unit; the data acquisition unit is used to acquire the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data of the patient; the information identification unit is used to analyze the intracranial aneurysm image data, the intracranial aneurysm auxiliary information data and the intracranial aneurysm basic information analysis model to obtain the intracranial aneurysm diagnosis information results and the intracranial aneurysm diagnosis information confidence; The special intracranial aneurysm information processing module includes a special information identification unit; the special information identification unit is used to input the intracranial aneurysm image data and the intracranial aneurysm auxiliary information data into the intracranial aneurysm special information analysis model for analysis to obtain the intracranial aneurysm special diagnosis information results when the intracranial aneurysm diagnosis information confidence level is lower than the preset information confidence level threshold; otherwise, no operation is performed; the intracranial aneurysm special information analysis model further screens and identifies key image features through secondary feature processing and special feature recognition technology; and subsequent operations are performed on the patient based on the intracranial aneurysm diagnosis information results and / or the intracranial aneurysm special diagnosis information results.

Images