Cerebrovascular disease risk prediction method and device based on distributed biobank

Abstract

Embodiments of the present application disclose a cerebrovascular disease risk prediction method and device based on a distributed biological sample library. The embodiments of the present application can determine at least one target prediction end matched with a clinical feature vector from a plurality of candidate prediction ends based on a first collaborative prediction request sent by a prediction request end and local sample data features of the plurality of candidate prediction ends, send a second collaborative prediction request to each target prediction end, receive a second initial cerebrovascular disease risk probability and a corresponding second prediction confidence returned by each target prediction end, obtain a comprehensive prediction confidence by weighted fusion of a first prediction confidence and each second prediction confidence, and obtain a target cerebrovascular disease risk probability by weighted fusion of a first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability, and send the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end. Thus, the embodiments of the present application improve the accuracy and reliability of cerebrovascular disease risk prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical artificial intelligence, in particular to a cerebrovascular disease risk prediction method and device based on a distributed biological sample library. BACKGROUND

[0002] Cerebrovascular diseases (such as ischemic stroke, cerebral hemorrhage, etc.) have the characteristics of high incidence, high disability rate and high mortality. Early risk prediction is of great significance for disease prevention and intervention. At present, the commonly used cerebrovascular disease risk prediction models (such as Framingham stroke risk score, QRISK, etc.) are mostly developed based on the data of patients treated in a hospital or research institution in the past. These data are usually collected retrospectively after the occurrence of the disease, only reflecting the health characteristics of a specific region, a specific population (such as a certain gender, age group or race), and lack sufficient diversity and representativeness.

[0003] With the development of artificial intelligence technology, some medical institutions have begun to use the electronic health records (such as physical examination results, medical history, test indicators, etc.) accumulated in the hospital to train localized machine learning models for individual risk assessment of patients. However, due to the obvious limitations of the patient samples that a single medical institution can obtain in terms of population coverage, clinical variable dimension and long-term follow-up integrity, the trained model is often difficult to adapt to external patients from different regions and different health backgrounds. When facing individuals with large differences in clinical characteristics and local training data, the model is prone to bias, resulting in unstable prediction results, decreased accuracy, and even misleading risk judgments. SUMMARY

[0004] The embodiments of the present application provide a cerebrovascular disease risk prediction method and device based on a distributed biological sample library, which overcomes the problem of insufficient generalization ability of single-center models due to sample limitations, and improves the accuracy and reliability of cerebrovascular disease risk prediction by safely cooperating with external prediction resources.

[0005] The embodiments of the present application provide a cerebrovascular disease risk prediction method based on a distributed biological sample library, the method comprising:

[0006] receiving a first collaborative prediction request sent by a prediction request end, the first collaborative prediction request carrying a clinical feature vector, a first initial cerebrovascular disease risk probability and its corresponding first prediction confidence, the clinical feature vector being extracted by the prediction request end based on the clinical feature data of a to-be-diagnosed object, and the first initial cerebrovascular disease risk probability and its corresponding first prediction confidence being generated by the prediction request end based on the clinical feature vector using a cerebrovascular disease risk prediction model deployed locally by the prediction request end;

[0007] determine, from the plurality of candidate prediction ends, at least one target prediction end matching the clinical feature vector based on the first collaborative prediction request and local sample data features of the plurality of candidate prediction ends, the local sample data features being used to represent data attributes of a local biological sample library used by the corresponding candidate prediction end when training the brain vascular disease risk prediction model locally deployed by the corresponding candidate prediction end;

[0008] send a second collaborative prediction request to each target prediction end, the second collaborative prediction request carrying the clinical feature vector, so that each target prediction end generates a corresponding second initial brain vascular disease risk probability and a corresponding second prediction confidence based on the clinical feature vector using the brain vascular disease risk prediction model locally deployed by the corresponding target prediction end;

[0009] receive the second initial brain vascular disease risk probability and the corresponding second prediction confidence returned by each target prediction end;

[0010] weight and fuse the first prediction confidence and each second prediction confidence to obtain a comprehensive prediction confidence, and weight and fuse the first initial brain vascular disease risk probability and each second initial brain vascular disease risk probability to obtain a target brain vascular disease risk probability;

[0011] send the target brain vascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end.

[0012] The embodiment of the application also provides a brain vascular disease risk prediction device based on a distributed biological sample library, and the device comprises:

[0013] a request receiving unit configured to receive a first collaborative prediction request sent by a prediction request end, the first collaborative prediction request carrying a clinical feature vector, a first initial brain vascular disease risk probability, and a corresponding first prediction confidence, the clinical feature vector being extracted by the prediction request end based on clinical feature data of a subject to be diagnosed, and the first initial brain vascular disease risk probability and the corresponding first prediction confidence being generated by the prediction request end based on the clinical feature vector using a brain vascular disease risk prediction model locally deployed by the prediction request end;

[0014] a matching unit configured to determine, from a plurality of candidate prediction ends, at least one target prediction end matching the clinical feature vector based on the first collaborative prediction request and local sample data features of the plurality of candidate prediction ends, the local sample data features being used to represent data attributes of a local biological sample library used by the corresponding candidate prediction end when training the brain vascular disease risk prediction model locally deployed by the corresponding candidate prediction end;

[0015] The request sending unit is configured to send a second collaborative prediction request to each target prediction terminal, and the second collaborative prediction request carries a clinical feature vector, so that each target prediction terminal generates a corresponding second initial cerebrovascular disease risk probability and a corresponding second prediction confidence based on the clinical feature vector and using a locally deployed cerebrovascular disease risk prediction model thereof;

[0016] The data receiving unit is configured to receive the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence returned by each target prediction terminal.

[0017] The weighted fusion unit is configured to fuse the first prediction confidence and each second prediction confidence to obtain a comprehensive prediction confidence, and to perform weighted fusion on the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability to obtain a target cerebrovascular disease risk probability.

[0018] The data sending unit is configured to send the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request terminal.

[0019] The embodiment of the present application also provides an electronic device, which comprises a processor and a memory, and the memory stores a plurality of instructions; the processor loads the instructions from the memory to execute the steps in any of the cerebrovascular disease risk prediction methods based on a distributed biological sample library provided by the embodiment of the present application.

[0020] The embodiment of the present application also provides a computer readable storage medium, which stores a plurality of instructions, and the instructions are adapted to be loaded by a processor to execute the steps in any of the cerebrovascular disease risk prediction methods based on a distributed biological sample library provided by the embodiment of the present application.

[0021] The embodiment of the present application also provides a computer program product, which comprises a computer program / instruction, and the computer program / instruction is executed by a processor to implement the steps in any of the cerebrovascular disease risk prediction methods based on a distributed biological sample library provided by the embodiment of the present application.

[0022] In the present application, after receiving the first collaborative prediction request sent by the prediction request terminal, at least one target prediction terminal that is most matched with the clinical feature vector is selected based on the clinical feature vector carried in the first collaborative prediction request and in combination with the local sample data features of the plurality of candidate prediction terminals. Since the local sample data features are used to represent the data properties of the biological sample library used by the corresponding candidate prediction terminal when training the locally deployed cerebrovascular disease risk prediction model thereof, through this matching mechanism, the training data distribution of the selected target prediction terminal has higher similarity with the clinical features of the object to be diagnosed, thereby ensuring that the cerebrovascular disease risk prediction model locally deployed by the target prediction terminal has stronger applicability and reliability in the current prediction task.

[0023] Further, after sending the second collaborative prediction request to each target prediction terminal, the target prediction terminals can generate a corresponding second initial cerebrovascular disease risk probability and a corresponding second prediction confidence based on the clinical feature vector and using the cerebrovascular disease risk prediction model locally deployed by the target prediction terminal, respectively. Since the target prediction terminal is selected by the aforementioned matching mechanism, the model output is more suitable for the clinical background of the subject to be diagnosed, and thus the prediction result is more targeted and reliable.

[0024] Subsequently, after receiving the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence returned by each target prediction terminal, a comprehensive prediction confidence is obtained by weighted fusion of the first prediction confidence and each second prediction confidence, and a target cerebrovascular disease risk probability is obtained by weighted fusion of the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability. This fusion mechanism effectively integrates the prediction opinions from the cerebrovascular disease risk prediction models locally deployed by multiple target prediction terminals, and realizes high-quality and high-confidence collaborative decision-making under the premise of fully respecting the data sovereignty and model independence of each target prediction terminal. Finally, the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence are returned to the prediction request terminal, so that the prediction request terminal can obtain a target cerebrovascular disease risk probability and a corresponding comprehensive prediction confidence with high accuracy and high reliability, thereby supporting more scientific and reliable clinical decision-making and significantly improving the effectiveness and safety of early warning of cerebrovascular diseases.

[0025] Thus, the problem of insufficient generalization ability caused by sample limitations of single-center models is overcome, and the accuracy and reliability of cerebrovascular disease risk prediction are improved by safely collaborating with external prediction resources. BRIEF DESCRIPTION OF DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0027] Figure 1 is a flowchart of a cerebrovascular disease risk prediction method based on a distributed biological sample library provided by an embodiment of the present application;

[0028] Figure 2 is a structural diagram of a cerebrovascular disease risk prediction device based on a distributed biological sample library provided by an embodiment of the present application. DETAILED DESCRIPTION

[0029] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative work fall within the scope of the present application.

[0030] The embodiment of the present application provides a cerebrovascular disease risk prediction method and device based on a distributed biological sample library.

[0031] The cerebrovascular disease risk prediction device based on the distributed biological sample library can be integrated in an electronic device, which can be a terminal, a server or the like. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a notebook computer or a personal computer (PC) or the like. The server can be a single server or a server cluster composed of multiple servers.

[0032] In some embodiments, the cerebrovascular disease risk prediction device based on the distributed biological sample library can also be integrated in multiple electronic devices, for example, the cerebrovascular disease risk prediction device based on the distributed biological sample library can be integrated in multiple servers to implement the cerebrovascular disease risk prediction method based on the distributed biological sample library of the present application.

[0033] In some embodiments, the server can also be implemented in the form of a terminal.

[0034] It can be understood that in the specific embodiments of the present application, the clinical feature data of the subject to be diagnosed, the local biological sample library and other related data are involved. When the embodiments of the present application are applied to specific products or technologies, the user's permission or consent is required, and the collection, use and processing of the related data need to comply with the relevant laws, regulations and standards of the country and region.

[0035] The following will be described in detail. It should be noted that the serial numbers of the following embodiments do not limit the preferred order of the embodiments.

[0036] In the embodiment, a cerebrovascular disease risk prediction method based on a distributed biological sample library is provided. The method is applicable to a center node with coordination capability, which can be a server, a cloud server or a coordination device deployed in a trusted computing environment, etc. Figure 1 For example, the specific process of the cerebrovascular disease risk prediction method based on the distributed biological sample library can be as follows:

[0037] 101. receive a first collaborative prediction request sent by a prediction requestor, the first collaborative prediction request carrying a clinical feature vector, a first initial cerebrovascular disease risk probability and its corresponding first prediction confidence, the clinical feature vector being extracted by the prediction requestor based on clinical feature data of a subject to be diagnosed, and the first initial cerebrovascular disease risk probability and its corresponding first prediction confidence being generated by the prediction requestor based on the clinical feature vector using a cerebrovascular disease risk prediction model deployed locally by the prediction requestor.

[0038] wherein the prediction requestor refers to a party initiating a collaborative prediction task, usually a medical institution or a computing node that performs preliminary screening or admission of the subject to be diagnosed, which has deployed a local cerebrovascular disease risk prediction model.

[0039] The first collaborative prediction request is a collaborative prediction trigger message sent by the prediction requestor to the center node, containing at least the clinical feature vector, the first initial cerebrovascular disease risk probability and its corresponding first prediction confidence, and is used to request the center node to coordinate multiple external predictionors to jointly participate in the cerebrovascular disease risk assessment of the subject to be diagnosed.

[0040] The subject to be diagnosed refers to an individual who needs to be assessed for cerebrovascular disease risk, such as a patient or a healthy person undergoing physical examination.

[0041] The clinical feature data refers to structured or semi-structured medical data related to the subject to be diagnosed and useful for cerebrovascular disease risk assessment, including but not limited to age, gender, blood pressure, blood glucose, blood lipids, smoking history, past medical history, family history, imaging indicators, laboratory test results, etc.

[0042] The clinical feature vector is a numerical vector representation extracted by the prediction requestor based on the clinical feature data through feature engineering or embedding coding, etc., as input of the cerebrovascular disease risk prediction model.

[0043] The cerebrovascular disease risk prediction model is a machine learning or statistical model deployed on each predictionor (including the prediction requestor and candidate predictionors), which is used to output the risk probability of the subject to be diagnosed for cerebrovascular disease (such as ischemic stroke, cerebral hemorrhage, etc.) within a specific time window in the future according to the input clinical feature vector. The cerebrovascular disease risk prediction model can be any form such as logistic regression, random forest, XGBoost, neural network, etc., and the model structures of the predictionors can be heterogeneous.

[0044] The first initial cerebrovascular disease risk probability is a preliminary risk estimate value calculated by the prediction requestor based on the clinical feature vector using the cerebrovascular disease risk prediction model deployed locally by the prediction requestor for the subject to be diagnosed to have cerebrovascular disease.

[0045] The first prediction confidence is a reliability evaluation value of the first initial cerebrovascular disease risk probability output by the prediction request end, reflecting the credibility of the prediction result under the current input condition. The confidence can be generated based on model uncertainty estimation (such as Bayesian method), calibration curve, historical performance statistics, etc.

[0046] 102. Based on the first collaborative prediction request and the local sample data features of the plurality of candidate prediction ends, at least one target prediction end matching the clinical feature vector is determined from the plurality of candidate prediction ends, and the local sample data features are used to represent the data properties of the local biological sample library used by the corresponding candidate prediction end when training the locally deployed cerebrovascular disease risk prediction model.

[0047] Wherein, the candidate prediction end refers to other participating nodes with local cerebrovascular disease risk prediction capability except the prediction request end after receiving the first collaborative prediction request. These nodes each deploy an independently trained cerebrovascular disease risk prediction model and have a corresponding local biological sample library. In the collaborative prediction process, the center node excludes the prediction request end itself from all available prediction ends, and the remaining prediction ends are used as candidate objects for subsequent matching and screening.

[0048] The local biological sample library refers to the biomedical data set owned by each prediction end (including the prediction request end and the candidate prediction end) for training its local cerebrovascular disease risk prediction model. The sample library usually contains clinical feature data, follow-up records, image data, laboratory test results, genetic information, etc. of historical patients, and is associated with a specific medical institution or research team, with inherent properties such as regional, population specificity and data collection standards.

[0049] Data properties are metadata information used to describe the overall characteristics of the local biological sample library, reflecting the key quality and distribution characteristics of the sample library in the construction and use process. Typical data properties include but are not limited to: population composition (such as age, gender, race distribution), geographical coverage, disease spectrum characteristics, clinical variable dimension integrity, data missing rate, follow-up length, annotation consistency, sample size, etc.

[0050] The local sample data features are a structured or vectorized representation of the data properties of the local biological sample library, which are extracted by each candidate prediction end and registered to the center node. The local sample data features are used to represent the data properties of the local biological sample library relied on by the corresponding candidate prediction end when training the locally deployed cerebrovascular disease risk prediction model, and are the basic input for matching calculation by the center node.

[0051] The target prediction end refers to one or more candidate prediction ends that are most matched with the current object to be diagnosed after the central node performs similarity evaluation on the basis of the clinical feature vector and the local sample data features of each candidate prediction end. The local biological sample library of these target prediction ends is highly similar to the object to be diagnosed in terms of population distribution or clinical feature coverage, and therefore the brain vascular disease risk prediction model deployed locally by the target prediction ends is expected to have higher applicability and prediction reliability in the current prediction task.

[0052] In some embodiments, through the intelligent screening mechanism based on data distribution similarity, it is ensured that the external models participating in collaborative prediction are highly adapted to the object to be diagnosed in terms of training data distribution, effectively avoiding prediction deviation caused by domain shift. Compared with the traditional way of randomly selecting or fixed partners, the present scheme significantly improves the relevance and reliability of external prediction results, providing high-quality input for subsequent weighted fusion, thereby enhancing the accuracy, generalization ability and clinical applicability of brain vascular disease risk assessment as a whole. Based on the first collaborative prediction request and the local sample data features of the multiple candidate prediction ends, at least one target prediction end that matches the clinical feature vector is determined from the multiple candidate prediction ends, including:

[0053] Based on the first collaborative prediction request, the prediction ends other than the prediction request end among the multiple prediction ends are determined as candidate prediction ends;

[0054] The similarity between the local sample data features of each candidate prediction end and the clinical feature vector is determined.

[0055] The multiple candidate prediction ends are sorted according to the similarity, and the top N candidate prediction ends are selected as target prediction ends, where N is a positive integer.

[0056] The prediction end refers to a medical institution or computing node that has deployed a local brain vascular disease risk prediction model and accessed the collaborative prediction network, such as a certain first-class hospital, regional medical center or data platform of a research institution. Each prediction end has an independent local biological sample library, and trains its exclusive brain vascular disease risk prediction model based on the sample library.

[0057] The similarity is used to measure the matching degree between the clinical feature distribution of the object to be diagnosed and the overall data distribution of the local sample library of a certain candidate prediction end. In specific implementation, the clinical feature vector can be first mapped in the feature space (such as through a pre-trained medical embedding model or principal component analysis) to obtain a feature embedding representation with unified dimensions; at the same time, the local sample data features (such as statistical indicators such as median age of the population, prevalence rate of hypertension, and blood lipid detection coverage) of each candidate prediction end are converted into sample library embedding representations with the same dimensions; then the cosine similarity, Euclidean distance or correlation coefficient between the two are calculated as the matching basis.

[0058] For example, if the subject to be diagnosed is a 65-year-old male patient with a history of diabetes living in a northern city, the clinical feature vector after mapping reflects characteristics such as "old age, metabolic abnormalities, and cold region". The local biological sample library of the candidate prediction end A is mainly from a young population in the south, and the sample library embedding is biased towards "young age and no chronic disease". The sample library of the candidate prediction end B is from a large stroke registry in the north, containing a large number of elderly diabetic patients. At this time, it is calculated that the similarity between the sample library embedding of the candidate prediction end B and the clinical feature vector of the subject to be diagnosed is significantly higher than that of the candidate prediction end A, so the candidate prediction end B is selected as the target prediction end.

[0059] Through the intelligent screening mechanism based on data distribution similarity, it is ensured that the external model participating in collaborative prediction is highly adapted to the training data distribution of the subject to be diagnosed, effectively avoiding the prediction deviation caused by domain shift. Compared with the traditional way of randomly selecting or fixed partners, the present scheme significantly improves the relevance and credibility of the external prediction result, provides high-quality input for subsequent weighted fusion, and thus enhances the accuracy, generalization ability and clinical applicability of the cerebral vascular disease risk assessment as a whole.

[0060] In some embodiments, through the embedding alignment mechanism, precise matching across granularities (individual vs. group) and modalities (clinical records vs. metadata) is achieved, significantly improving the scientificity and rationality of target prediction end screening. This not only enhances the pertinence of collaborative prediction, but also lays a foundation for high-quality input for subsequent high-confidence fusion, ultimately effectively improving the overall accuracy and robustness of cerebral vascular disease risk assessment. The similarity between the local sample data features of each candidate prediction end and the clinical feature vector is determined, including:

[0061] Mapping the clinical feature vector to a feature space to obtain a feature embedding representation;

[0062] Obtaining the local sample data features corresponding to each candidate prediction end, and converting the local sample data features into sample library embedding representations of the same dimension as the feature embedding representation;

[0063] Calculating the similarity between the feature embedding representation and each sample library embedding representation.

[0064] The feature embedding representation refers to a dense vector obtained by projecting the clinical feature vector of the subject to be diagnosed into a low-dimensional continuous vector space through a pre-set mapping function (such as a fully connected neural network, principal component analysis, or medical knowledge graph embedding model), which is used to more effectively represent the comprehensive clinical state and potential risk patterns of the subject to be diagnosed.

[0065] Sample library embedding representation refers to converting the local sample data features of a candidate prediction end (such as the average age of patients in the institution's biobank, the prevalence of hypertension, the completeness rate of blood lipid testing, the entropy value of regional distribution, and other statistical metadata) into a vector form with the same or compatible dimensions in the same vector space as the above feature embedding representation through the same or compatible mapping mechanism, so that the two can be compared in a unified semantic space.

[0066] Understandably, by mapping the clinical feature vector of the subject to be diagnosed to the local sample data features corresponding to the candidate prediction model into the same embedding space, it is possible to quantify, in a numerical and computable way, whether "the patient belongs to the typical population of a certain institution's sample database". Based on this, the similarity between the two can be calculated by using methods such as cosine similarity, Euclidean distance, or inner product, which can objectively reflect the applicability potential of the target prediction model for the patient.

[0067] 103. Send a second collaborative prediction request to each target prediction terminal. The second collaborative prediction request carries a clinical feature vector so that each target prediction terminal can generate a corresponding second initial cerebrovascular disease risk probability and its corresponding second prediction confidence based on the clinical feature vector and its locally deployed cerebrovascular disease risk prediction model.

[0068] The second collaborative prediction request is a collaborative reasoning trigger message sent by the central node to the selected target prediction end. It is used to request the target prediction end to perform local risk prediction based on the clinical feature vector of the current object to be diagnosed. The request includes at least the clinical feature vector and optionally also auxiliary information such as task identifier, encryption key or timestamp.

[0069] In some embodiments, the clinical feature vector is encrypted before transmission to protect patient privacy.

[0070] The second initial cerebrovascular disease risk probability refers to the preliminary risk estimate of the subject's potential cerebrovascular disease (such as ischemic stroke, cerebral hemorrhage, etc.) calculated by any target prediction terminal after receiving the second collaborative prediction request, using its locally deployed cerebrovascular disease risk prediction model and based on the clinical feature vectors it carries. This probability value is generated independently and reflects the risk judgment of the target prediction terminal under its own training data distribution.

[0071] The second prediction confidence score refers to the reliability assessment value of the second initial cerebrovascular disease risk probability output by the corresponding target prediction end, used to characterize the credibility of the prediction result under the current input conditions. The confidence score can be generated based on various methods such as internal model uncertainty (such as the variance of a Bayesian neural network), calibrated probability confidence intervals, accuracy-confidence consistency on the historical validation set, or sample library quality weighting. The higher the value, the more reliable the prediction.

[0072] In some embodiments, the core problem of "data availability invisibility" in multi-center medical AI collaboration is effectively solved, the design enhances the compliance and user trust of the system, and a second collaborative prediction request is sent to each target prediction end, including:

[0073] The clinical feature vector is encrypted to obtain an encrypted clinical feature vector.

[0074] The second collaborative prediction request is generated according to the encrypted clinical feature vector.

[0075] The second collaborative prediction request is sent to each target prediction end.

[0076] The encrypted clinical feature vector refers to the ciphertext form data generated by protecting the clinical feature vector by using a cryptography method (such as homomorphic encryption, symmetric encryption, asymmetric encryption, or secure multi-party computation protocol).

[0077] It can be understood that although the clinical feature vector is a structured representation extracted by the prediction request end based on the clinical feature data of the object to be diagnosed, it still contains a large amount of identifiable or sensitive personal health information (such as age, medical history, test indicators, etc.). If transmitted in plaintext form in the network, there is a risk of interception, misuse, or association restoration, which may lead to privacy leakage of the object to be diagnosed. Therefore, the clinical feature vector is encrypted before being sent to the target prediction end to generate an encrypted clinical feature vector. The encryption mechanism ensures that the target prediction end can only use the data under authorized conditions, for example, by supporting homomorphic encryption to directly perform reasoning on the ciphertext, or by restoring the plaintext in a trusted execution environment (such as TEE) with secure decryption permission to make predictions, and the center node only contacts and forwards the encrypted clinical feature vector in the entire collaborative prediction process, without accessing, decrypting, or storing any original plaintext patient data. Thus, the exposure path of sensitive information is effectively isolated, ensuring the normal implementation of multi-center collaborative prediction while significantly improving the privacy security of the entire life cycle of data. This mechanism effectively meets the core security requirements of "data availability invisibility" in cross-institutional collaboration in the medical and health field, providing reliable technical support for the compliance deployment and large-scale application of the cerebrovascular disease risk prediction system.

[0078] 104、Receiving the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence returned by each target prediction end.

[0079] 105、Weighted fusion of the first prediction confidence and each second prediction confidence to obtain a comprehensive prediction confidence, and weighted fusion of the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability to obtain a target cerebrovascular disease risk probability.

[0080] The comprehensive prediction confidence refers to a unified confidence index generated by weighting and fusing the original confidence (i.e., the first prediction confidence and the second prediction confidence) provided by the prediction request end and each target prediction end, in combination with the data quality, matching degree or reliability weight of the corresponding model. The value is used to quantify the overall confidence degree of the final risk assessment result.

[0081] The target cerebrovascular disease risk probability refers to the final cerebrovascular disease risk estimate value generated by fusing the opinions of the local and multiple external high-adaptation prediction ends, reflecting the comprehensive probability of the subject to be diagnosed to have a cerebrovascular event (such as ischemic stroke, cerebral hemorrhage, etc.) within a specific time window in the future. The probability is not a simple average, but a result of dynamic weighting according to the prediction confidence of each participant and the quality of the sample library, with stronger generalization ability and clinical representativeness.

[0082] It can be understood that, due to the differences in model training data distribution, sample quality and applicable scenarios of different prediction ends, the reliability of the initial risk probability and confidence output by them is also different. If an equal-weighted average or a fusion method that ignores confidence is used, it is easy to be disturbed by low-quality or mismatched nodes, resulting in deviation of the final result. By weighting and fusing with confidence as the weight, the prediction opinions with high confidence and high adaptation dominate in the final decision, so as to effectively suppress the influence of noise while retaining the diversity of multiple sources, and improve the accuracy, robustness and clinical interpretability of the overall prediction.

[0083] In some embodiments, the interference of low-quality prediction ends on the final result is effectively avoided, and the accuracy and robustness of the fusion result are significantly improved; at the same time, the comprehensive prediction confidence not only reflects the model confidence, but also embeds data quality information, so that clinicians can more comprehensively evaluate the reliable basis for risk judgment, thereby supporting safer and more individualized intervention decisions. The first prediction confidence and each second prediction confidence are weighted and fused to obtain a comprehensive prediction confidence, and the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability are weighted and fused to obtain a target cerebrovascular disease risk probability, comprising:

[0084] Obtaining a first sample quality score corresponding to the local sample data features of the prediction request end, and a second sample quality score corresponding to the local sample data features of each target prediction end;

[0085] Multiplying the first sample quality score by the first prediction confidence to obtain a first compound weight, and multiplying each second sample quality score by the corresponding second prediction confidence to obtain a plurality of second compound weights;

[0086] Normalizing the first compound weight and each second compound weight to obtain a first normalized weight and a plurality of second normalized weights;

[0087] weighting the first initial cerebrovascular disease risk probability based on the first normalization weight to obtain a first weighted cerebrovascular disease risk probability, and weighting the second initial cerebrovascular disease risk probability corresponding to each second normalization weight to obtain a plurality of second weighted cerebrovascular disease risk probabilities;

[0088] fusing the first weighted cerebrovascular disease risk probability and the plurality of second weighted cerebrovascular disease risk probabilities to obtain a target cerebrovascular disease risk probability;

[0089] weighting and summing the first prediction confidence and the second prediction confidence based on the first normalization weight and the second normalization weight to obtain a comprehensive prediction confidence.

[0090] The first sample quality score is used to quantify the overall data quality level of the local biological sample library relied on by the prediction request end, and is calculated based on a plurality of preset quality evaluation dimensions (such as sample size, population diversity, clinical variable integrity, follow-up time length, annotation consistency, etc.) to reflect the reliability and representativeness of the model training data of the end.

[0091] The second sample quality score is used to quantify the data quality of the respective local biological sample library of the target prediction end, and is an important basis for evaluating the credibility of the prediction result.

[0092] The first composite weight is obtained by multiplying the first sample quality score of the prediction request end by the first prediction confidence output by the prediction request end, and comprehensively reflects the dual reliability of "data quality x model confidence" of the end in the current task.

[0093] The second composite weight is obtained by multiplying the second sample quality score of each target prediction end by the second prediction confidence corresponding to the target prediction end, and is used to represent the comprehensive credible contribution of the corresponding target prediction end in this collaboration.

[0094] The first normalization weight / second normalization weight is a weight coefficient obtained by normalizing (such as softmax or linear normalization) all composite weights (including the first normalization weight and the second normalization weight), ensuring that the sum of the weighting coefficients of each participant is 1, facilitating subsequent probability fusion and confidence aggregation.

[0095] The first weighted cerebrovascular disease risk probability refers to a weighted component obtained by multiplying the first initial cerebrovascular disease risk probability by the corresponding normalization weight, representing the contribution value of the prediction request end in the final result.

[0096] The second weighted cerebrovascular disease risk probability refers to a weighted component obtained by multiplying the second initial cerebrovascular disease risk probability of each target prediction end by the corresponding normalization weight, representing the contribution value of each external prediction end in the final result.

[0097] It can be understood that the traditional multi-model fusion method usually only relies on the confidence of the model output or simple average, ignoring the quality difference of the training data behind the model. In the medical scene, a high-confidence prediction may be derived from a small sample, a single population, or a low-quality sample library missing key variables, and its actual reliability may be much lower than expected. The present application realizes the intelligent fusion strategy of "high-quality data + high-confidence output = high weight adoption" by introducing the joint weighting mechanism of sample quality score and prediction confidence.

[0098] In some embodiments, the core problem of "model available but data not comparable" in multi-center medical AI collaboration is effectively solved, and intelligent collaboration based on data quality perception is realized without sharing raw data, significantly improving the accuracy, robustness and clinical landing feasibility of cerebrovascular disease risk prediction. Before obtaining the first sample quality score corresponding to the local sample data features of the prediction request end and the second sample quality score corresponding to the local sample data features of each target prediction end, it further comprises:

[0099] Based on the plurality of preset quality evaluation dimensions, the actual feature measurement value corresponding to each preset quality evaluation dimension is extracted from the local sample data features of each prediction end;

[0100] For each prediction end, based on the mapping relationship between the feature measurement value of the preset quality evaluation dimension and the preset score, the evaluation score of the actual feature measurement value corresponding to each preset quality evaluation dimension is determined;

[0101] The evaluation scores of the actual feature measurement values corresponding to each preset quality evaluation dimension are summed to obtain the sample quality score;

[0102] Obtaining the first sample quality score corresponding to the local sample data features of the prediction request end and the second sample quality score corresponding to the local sample data features of each target prediction end comprises:

[0103] Obtaining the first sample quality score corresponding to the local sample data features of the prediction request end and the second sample quality score corresponding to the local sample data features of each target prediction end from the plurality of sample quality scores.

[0104] Among them, the preset quality evaluation dimension refers to a set of standardized index dimensions for systematically evaluating the data quality of the local biological sample library, which is pre-set by those skilled in the art according to medical research specifications (such as STROBE, TRIPOD, etc.). Typical dimensions include but are not limited to: sample size, population diversity (such as age / gender / region distribution entropy), clinical variable coverage completeness, key risk factor missing rate, follow-up length, event annotation accuracy, data collection consistency, etc.

[0105] The actual feature metric value refers to a specific quantitative value calculated for the local sample data features of a certain prediction end under a specific preset quality evaluation dimension. For example, under the dimension of "clinical variable coverage completeness", the actual feature metric value can be "9 out of 10 core stroke risk factors have complete records, with a completeness rate of 90%".

[0106] The mapping relationship between the feature metric value of the preset quality evaluation dimension and the preset score refers to a pre-established scoring rule or function for converting the actual feature metric value of a certain dimension into a corresponding evaluation score. The mapping relationship can be a piecewise linear function, a lookup table method, a threshold classification (e.g., a completeness rate ≥ 90% gets 5 points, 80%~90% gets 3 points, and < 80% gets 1 point), or a machine learning calibration model, the purpose of which is to convert objective data quality into a comparable numerical score.

[0107] The sample quality score refers to a comprehensive quantitative score of the overall data quality of a local biological sample library of a prediction end, which is obtained by summing (or weighted fusion) the evaluation scores of each preset quality evaluation dimension. The higher the score, the more reliable and representative the training data of the prediction end, and the more reliable the prediction results output by the local model.

[0108] It can be understood that biological sample libraries of different prediction ends (medical institutions) have significant differences in construction standards, data governance levels, and clinical record specifications. If all prediction ends are treated equally in collaborative prediction without considering their underlying data quality, it may lead to misleading effects of the model trained by low-quality sample libraries on the final results. The present application realizes objective, quantitative, and interpretable credibility evaluation of the prediction capabilities of each participant by introducing a sample quality score mechanism based on multi-dimensional quality evaluation.

[0109] In some embodiments, by introducing a dynamic participation mechanism based on timeout detection, only high-quality prediction results that are actually available and returned in time are fused under the premise of ensuring prediction timeliness, effectively avoiding system congestion or result distortion caused by node failure or communication abnormalities, significantly improving the robustness, real-time performance, and fault tolerance of the collaborative prediction system, ensuring stable output of high-credibility cerebrovascular disease risk assessment results in a heterogeneous and unreliable multi-center environment, and meeting the actual needs of clinical "fast, reliable, and uninterrupted" intelligent auxiliary decision-making; receiving the second initial cerebrovascular disease risk probability returned by each target prediction end and the corresponding second prediction confidence, including:

[0110] detecting the response state of each target prediction end;

[0111] If at least one target prediction end fails to return the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence within the preset response time, the unresponsive target prediction end will be removed from the participation list of the current collaborative prediction task.

[0112] Receive the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence level returned by the remaining target prediction terminals that have responded.

[0113] The response status refers to the status indicator of whether the target prediction end has successfully completed local inference and returned a valid prediction result after receiving the second collaborative prediction request, including "responded" (successfully returned) or "not responded" (timeout, failure, or no return).

[0114] The preset response time refers to the maximum waiting time (e.g., 5 seconds, 10 seconds, etc.) set by the central node for this collaborative prediction task, used to determine whether the target prediction end has completed the calculation and feedback within a reasonable time. This time can be dynamically configured based on network conditions, model complexity, or historical performance.

[0115] Unresponsive target prediction ends refer to candidate target prediction ends that fail to return a valid second initial cerebrovascular disease risk probability and a second prediction confidence level to the central node within the preset response time. Their failure to respond may be due to reasons such as network interruption, insufficient computing resources, model abnormality or service unavailability.

[0116] The current collaborative prediction task participation list refers to the set of all target prediction ends selected in this prediction process. It is used to track the task execution status of each participant and serves as the input source for subsequent fusion calculations.

[0117] The target prediction end that has responded refers to the target prediction end that successfully returns a valid prediction result within the preset response time. Its output will be included in the weighted fusion stage and participate in the generation of the final risk assessment.

[0118] Understandably, in distributed medical collaboration scenarios spanning multiple institutions and regions, there are significant differences in the computing power, network stability, and system load of each target prediction endpoint, and some nodes may experience delays or failures. Forcing all nodes to respond will result in excessively high overall prediction latency, affecting clinical timeliness; while blindly using partially missing results for fusion may introduce bias or reduce confidence.

[0119] 106. Send the target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence level to the prediction request end.

[0120] In some embodiments, the clinical acceptability and decision support value of the cerebral vascular disease risk prediction system are significantly enhanced; at the same time, by highlighting key risk factors and their contribution, doctors can also assist patients in personalized health education or develop precise prevention strategies, promoting the transition from "passive treatment" to "active intervention", improving the overall public health prevention and control effectiveness, and sending the target cerebral vascular disease risk probability and its corresponding comprehensive prediction confidence to the prediction request end, including:

[0121] Generating risk explanation information, the risk explanation information including a number of clinical characteristics and their contribution degrees that have the greatest impact on the target cerebral vascular disease risk probability;

[0122] Sending the risk explanation information together with the target cerebral vascular disease risk probability and its corresponding comprehensive prediction confidence to the prediction request end.

[0123] Among them, the risk explanation information refers to the explainable auxiliary information for explaining the cause of the final cerebral vascular disease risk assessment result, aiming to reveal which clinical factors play a key role in the current high (or low) risk judgment, thereby enhancing the transparency and credibility of the model output.

[0124] Clinical characteristics refer to the original medical attributes of the subject to be diagnosed, such as age, systolic blood pressure, diabetes history, smoking status, total cholesterol level, previous stroke history, etc. These characteristics constitute the input basis for risk prediction.

[0125] Contribution degree refers to the quantitative influence degree of a certain clinical characteristic on the target cerebral vascular disease risk probability, which can be calculated by model explanation techniques such as SHAP value, LIME, gradient significance, feature ablation analysis, etc. For example, the contribution degree of the "age" feature is +0.15, indicating that this feature increases the risk probability by 15 percentage points.

[0126] It can be understood that although deep learning or ensemble models have high accuracy in cerebral vascular disease risk prediction, their "black box" characteristics often make it difficult for clinicians to trust or adopt the prediction results. Especially in high-risk warning scenarios, if the reason for determining high risk cannot be explained, the clinical applicability of AI systems will be greatly limited. The present application generates clinical explainable feedback automatically while returning the final prediction result, so that the prediction request end (such as a doctor's workstation or a health management platform) not only knows "how high the risk is", but also knows "where the risk comes from". This not only improves the transparency and traceability of the model, but also supports doctors to cross-verify or intervene and adjust the prediction results based on their professional knowledge.

[0127] In some embodiments, while ensuring the continuity of the prediction service, the risk of misjudgment caused by data missing or model mismatch is effectively prevented, the completeness of clinical information collection is promoted through data supplement prompts, a benign closed loop of "evaluation-feedback-optimization" is formed, and the practicality and credibility of brain vascular disease risk prediction in real medical environment are ultimately enhanced. The target brain vascular disease risk probability and its corresponding comprehensive prediction confidence are sent to the prediction request end, including:

[0128] When the comprehensive prediction confidence is less than the preset confidence threshold, an unreliable prompt corresponding to the target brain vascular disease risk probability is generated, and a data supplement prompt for the clinical feature data is generated;

[0129] The unreliable prompt and the data supplement prompt are sent to the prediction request end;

[0130] When the comprehensive prediction confidence is not less than the preset confidence threshold, the target brain vascular disease risk probability and its corresponding comprehensive prediction confidence are sent to the prediction request end.

[0131] The preset confidence threshold refers to a confidence threshold value (for example, 0.7, 0.8, etc.) preset by the system or configured by the user, which is used to determine whether the current collaborative prediction result meets the reliability standard of clinical use. The threshold value can be dynamically adjusted according to the application scenario (such as screening vs. diagnosis), disease severity or institutional strategy.

[0132] The unreliable prompt refers to the warning information automatically generated by the system when the comprehensive prediction confidence is lower than the preset confidence threshold, which is used to inform the prediction request end that "the current risk assessment result has insufficient reliability, and is not recommended for direct clinical decision making". For example: "the risk assessment result confidence is low, please refer to it carefully".

[0133] The data supplement prompt refers to further analyzing the missing or low-quality clinical feature dimensions when the confidence is insufficient, and suggesting to the prediction request end which key information should be supplemented to improve the prediction reliability. For example: "it is recommended to supplement the carotid ultrasound examination result" or "the blood lipid detection data is missing, which may affect the evaluation accuracy".

[0134] It can be understood that in real clinical scenarios, not all cases to be diagnosed can provide complete and high-quality clinical feature data. If a high-confidence risk value is forcibly output for cases with incomplete information or serious distribution deviation, misjudgment may occur, causing unnecessary anxiety or missed diagnosis risk. Traditional systems often "answer every question", lacking a self-evaluation and feedback mechanism for their own prediction reliability. The present application introduces a dynamic response strategy based on comprehensive prediction confidence, realizing an intelligent interaction mode of "high confidence output result, low confidence guide to complete". This not only avoids the blind use of low-quality prediction results, but also actively provides clear guidance for doctors or patients to improve data input.

[0135] In some embodiments, while sending the unreliable prompt and the data supplement prompt to the prediction request end, it can be decided according to a preset strategy whether to send the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence together.

[0136] Among them, the preset strategy decision can only send the unreliable prompt and the data supplement prompt to avoid interference of low confidence results on clinical judgment. Or send the target cerebrovascular disease risk probability and the comprehensive prediction confidence together for professional personnel to use in combination with context reference.

[0137] In summary, the embodiments of the present application overcome the problem of insufficient generalization ability caused by sample limitation of single center model, and improve the accuracy and reliability of cerebrovascular disease risk prediction by safely cooperating with external prediction resources.

[0138] In order to better implement the above method, the embodiments of the present application also provide a cerebrovascular disease risk prediction device based on distributed biological sample library. The cerebrovascular disease risk prediction device based on distributed biological sample library can be integrated in an electronic device. The electronic device can be a terminal, a server, etc. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a notebook computer, a personal computer, etc. The server can be a single server or a server cluster composed of multiple servers.

[0139] For example, in the present embodiment, a cerebrovascular disease risk prediction device based on distributed biological sample library is integrated in an electronic device as an example to illustrate the method of the present embodiment.

[0140] For example, as shown in Figure 2 The cerebrovascular disease risk prediction device based on distributed biological sample library can include a request receiving unit 201, a matching unit 202, a request sending unit 203, a data receiving unit 204, a weighted fusion unit 205, and a data sending unit 206, as follows:

[0141] (I) Request receiving unit 201.

[0142] The request receiving unit 201 is configured to receive a first cooperative prediction request sent by a prediction request end. The first cooperative prediction request carries a clinical feature vector, a first initial cerebrovascular disease risk probability and a corresponding first prediction confidence. The clinical feature vector is extracted by the prediction request end based on the clinical feature data of a to-be-diagnosed object. The first initial cerebrovascular disease risk probability and the corresponding first prediction confidence are generated by the prediction request end based on the clinical feature vector using a locally deployed cerebrovascular disease risk prediction model.

[0143] (II) Matching unit 202.

[0144] The matching unit 202 is configured to determine at least one target prediction end from the plurality of candidate prediction ends based on the first collaborative prediction request and local sample data features of the plurality of candidate prediction ends, wherein the local sample data features are used to represent data attributes of a local biological sample library used by the corresponding candidate prediction end when training the locally deployed cerebrovascular disease risk prediction model.

[0145] In some embodiments, determining at least one target prediction end from the plurality of candidate prediction ends based on the first collaborative prediction request and local sample data features of the plurality of candidate prediction ends comprises:

[0146] Based on the first collaborative prediction request, determining the prediction ends other than the prediction request end from the plurality of prediction ends as candidate prediction ends;

[0147] Determining the similarity between the local sample data features of each candidate prediction end and the clinical feature vector;

[0148] Ranking the plurality of candidate prediction ends according to the similarity, and selecting the top N candidate prediction ends as target prediction ends, wherein N is a positive integer.

[0149] In some embodiments, determining the similarity between the local sample data features of each candidate prediction end and the clinical feature vector comprises:

[0150] Performing feature space mapping on the clinical feature vector to obtain a feature embedding representation;

[0151] Obtaining the local sample data features corresponding to each candidate prediction end, and converting the local sample data features into sample library embedding representations with the same dimension as the feature embedding representation;

[0152] Calculating the similarity between the feature embedding representation and each sample library embedding representation.

[0153] (Three), the request sending unit 203.

[0154] The request sending unit 203 is configured to send a second collaborative prediction request to each target prediction end, wherein the second collaborative prediction request carries the clinical feature vector, so that each target prediction end generates a corresponding second initial cerebrovascular disease risk probability and a corresponding second prediction confidence based on the clinical feature vector and using the locally deployed cerebrovascular disease risk prediction model.

[0155] In some embodiments, sending the second collaborative prediction request to each target prediction end comprises:

[0156] Encrypting the clinical feature vector to obtain an encrypted clinical feature vector;

[0157] generating a second collaborative prediction request according to the encrypted clinical feature vector;

[0158] sending the second collaborative prediction request to each target prediction end.

[0159] (four), the data receiving unit 204.

[0160] The data receiving unit 204 is configured to receive the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence returned by each target prediction end.

[0161] In some embodiments, receiving the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence returned by each target prediction end comprises:

[0162] detecting the response state of each target prediction end;

[0163] If there is at least one target prediction end that does not return the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence within the preset response time, the target prediction end that does not respond is removed from the participation list of the current collaborative prediction task;

[0164] receiving the second initial cerebrovascular disease risk probability and the corresponding second prediction confidence returned by the remaining target prediction ends that have responded.

[0165] (five), the weighted fusion unit 205.

[0166] The weighted fusion unit 205 is configured to fuse the first prediction confidence and each second prediction confidence to obtain a comprehensive prediction confidence, and to perform weighted fusion on the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability to obtain a target cerebrovascular disease risk probability.

[0167] In some embodiments, the first prediction confidence and each second prediction confidence are fused to obtain a comprehensive prediction confidence, and the first initial cerebrovascular disease risk probability and each second initial cerebrovascular disease risk probability are weighted fused to obtain a target cerebrovascular disease risk probability, comprising:

[0168] obtaining a first sample quality score corresponding to the local sample data feature of the prediction request end, and a second sample quality score corresponding to the local sample data feature of each target prediction end;

[0169] multiplying the first sample quality score by the first prediction confidence to obtain a first compound weight, and multiplying each second sample quality score by the corresponding second prediction confidence to obtain a plurality of second compound weights;

[0170] normalizing the first compound weight and each second compound weight to obtain a first normalized weight and a plurality of second normalized weights;

[0171] weighting the first initial cerebrovascular disease risk probability based on the first normalization weight to obtain a first weighted cerebrovascular disease risk probability, and weighting the second initial cerebrovascular disease risk probability corresponding to each second normalization weight to obtain a plurality of second weighted cerebrovascular disease risk probabilities;

[0172] fusing the first weighted cerebrovascular disease risk probability and the plurality of second weighted cerebrovascular disease risk probabilities to obtain a target cerebrovascular disease risk probability;

[0173] weighting and summing the first prediction confidence and the second prediction confidence based on the first normalization weight and each second normalization weight to obtain a comprehensive prediction confidence.

[0174] In some embodiments, before obtaining the first sample quality score corresponding to the local sample data feature of the prediction request end and the second sample quality score corresponding to the local sample data feature of each target prediction end, the method further comprises:

[0175] based on a plurality of preset quality evaluation dimensions, extracting an actual feature measurement value corresponding to each preset quality evaluation dimension from the local sample data feature of each prediction end;

[0176] for each prediction end, determining an evaluation score of the actual feature measurement value corresponding to each preset quality evaluation dimension based on a mapping relationship between the feature measurement value of the preset quality evaluation dimension and the preset score;

[0177] summing the evaluation scores of the actual feature measurement values corresponding to each preset quality evaluation dimension to obtain a sample quality score;

[0178] obtaining the first sample quality score corresponding to the local sample data feature of the prediction request end and the second sample quality score corresponding to the local sample data feature of each target prediction end, comprises:

[0179] obtaining the first sample quality score corresponding to the local sample data feature of the prediction request end and the second sample quality score corresponding to the local sample data feature of each target prediction end from the plurality of sample quality scores.

[0180] (Six), the data sending unit 206.

[0181] The data sending unit 206 is configured to send the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end.

[0182] In some embodiments, sending the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end comprises:

[0183] generate risk explanation information, the risk explanation information including a number of clinical features and their contribution degrees that have the greatest impact on the target cerebrovascular disease risk probability;

[0184] send the risk explanation information, the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end.

[0185] In some embodiments, sending the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end includes:

[0186] When the comprehensive prediction confidence is less than a preset confidence threshold, generating an unreliable prompt for the target cerebrovascular disease risk probability and a data supplement prompt for the clinical feature data;

[0187] send the unreliable prompt and the data supplement prompt to the prediction request end;

[0188] When the comprehensive prediction confidence is not less than the preset confidence threshold, send the target cerebrovascular disease risk probability and the corresponding comprehensive prediction confidence to the prediction request end.

[0189] In specific implementation, the above various units can be implemented as independent entities, or can be combined as the same or several entities, and the specific implementation of the above various units can be referred to the method embodiments above, which will not be described here.

[0190] Therefore, the embodiments of the present application overcome the problem of insufficient generalization ability caused by sample limitation of single-center model, and improve the accuracy and reliability of cerebrovascular disease risk prediction by safely cooperating with external prediction resources.

[0191] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by related hardware controlled by the instructions, which can be stored in a computer readable storage medium and loaded and executed by a processor.

[0192] Therefore, the embodiments of the present application provide a computer readable storage medium, which stores a plurality of instructions capable of being loaded by a processor to execute the steps in any one of the cerebrovascular disease risk prediction methods based on distributed biological sample library provided by the embodiments of the present application.

[0193] The storage medium can include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0194] Due to the instructions stored in the storage medium, the steps in any of the trash rack cleaning path planning methods provided by the embodiments of the present application can be executed, thus the beneficial effects of any of the brain vascular disease risk prediction methods based on distributed biological sample library provided by the embodiments of the present application can be achieved, which will be described in detail in the foregoing embodiments, and will not be repeated here.

[0195] According to an aspect of the present application, a computer program product or computer program is provided, which includes computer programs / instructions stored in a computer readable storage medium. A processor of an electronic device reads the computer programs / instructions from the computer readable storage medium, and the processor executes the computer programs / instructions, so that the electronic device executes the method provided in the brain vascular disease risk prediction aspect based on distributed biological sample library provided in the foregoing embodiments.

[0196] The above describes in detail the method and device for predicting brain vascular disease risk based on distributed biological sample library provided by the embodiments of the present application, and the principles and implementation manners of the present application are described by applying specific examples. The above embodiment is only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manner and application range will be changed, and in conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for predicting the risk of cerebrovascular disease based on a distributed biobank, characterized in that, The method includes: The system receives a first collaborative prediction request sent by the prediction request terminal. The first collaborative prediction request carries a clinical feature vector, a first initial cerebrovascular disease risk probability and its corresponding first prediction confidence. The clinical feature vector is extracted by the prediction request terminal based on the clinical feature data of the object to be diagnosed. The first initial cerebrovascular disease risk probability and its corresponding first prediction confidence are generated by the prediction request terminal based on the clinical feature vector using its locally deployed cerebrovascular disease risk prediction model. Based on the first collaborative prediction request and the local sample data features of multiple candidate prediction ends, at least one target prediction end that matches the clinical feature vector is determined from the multiple candidate prediction ends. The local sample data features are used to characterize the data attributes of the local biobank used by the corresponding candidate prediction end when training its locally deployed cerebrovascular disease risk prediction model. A second collaborative prediction request is sent to each target prediction terminal. The second collaborative prediction request carries the clinical feature vector, so that each target prediction terminal can generate a corresponding second initial cerebrovascular disease risk probability and its corresponding second prediction confidence based on the clinical feature vector and its locally deployed cerebrovascular disease risk prediction model. Receive the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence level returned by each target prediction terminal; The first prediction confidence and each of the second prediction confidences are weighted and fused to obtain the comprehensive prediction confidence, and the first initial cerebrovascular disease risk probability and each of the second initial cerebrovascular disease risk probabilities are weighted and fused to obtain the target cerebrovascular disease risk probability. The target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence level are sent to the prediction request terminal.

2. The method as described in claim 1, characterized in that, The step of determining at least one target prediction endpoint that matches the clinical feature vector from multiple candidate prediction endpoints based on the local sample data features of the first collaborative prediction request and multiple candidate prediction endpoints includes: Based on the first collaborative prediction request, prediction ends other than the prediction request end among multiple prediction ends are determined as candidate prediction ends; Determine the similarity between the local sample data features of each candidate prediction end and the clinical feature vector; The multiple candidate prediction ends are sorted according to the similarity, and the top N candidate prediction ends are selected as the target prediction ends, where N is a positive integer.

3. The method as described in claim 2, characterized in that, Determining the similarity between the local sample data features of each candidate prediction endpoint and the clinical feature vector includes: The clinical feature vectors are mapped to a feature space to obtain a feature embedding representation; Obtain the local sample data features corresponding to each candidate prediction end, and convert the local sample data features into a sample library embedding representation of the same dimension as the feature embedding representation; Calculate the similarity between the feature embedding representation and each of the sample library embedding representations.

4. The method as described in claim 1, characterized in that, The weighted fusion of the first prediction confidence and each of the second prediction confidences to obtain a comprehensive prediction confidence, and the weighted fusion of the first initial cerebrovascular disease risk probability and each of the second initial cerebrovascular disease risk probabilities to obtain a target cerebrovascular disease risk probability, include: Obtain the first sample quality score corresponding to the local sample data features of the prediction request terminal, and the second sample quality score corresponding to the local sample data features of each target prediction terminal. The first sample quality score is multiplied by the first prediction confidence to obtain the first composite weight, and each second sample quality score is multiplied by its corresponding second prediction confidence to obtain multiple second composite weights. The first composite weight and each of the second composite weights are normalized to obtain the first normalized weight and multiple second normalized weights; The first initial cerebrovascular disease risk probability is weighted based on the first normalized weight to obtain the first weighted cerebrovascular disease risk probability, and the corresponding second initial cerebrovascular disease risk probability is weighted based on each second normalized weight to obtain multiple second weighted cerebrovascular disease risk probabilities. The first weighted cerebrovascular disease risk probability and the plurality of second weighted cerebrovascular disease risk probabilities are fused to obtain the target cerebrovascular disease risk probability; Based on the first normalized weight and each of the second normalized weights, the first prediction confidence and each of the second prediction confidence are weighted and summed to obtain the comprehensive prediction confidence.

5. The method as described in claim 4, characterized in that, Before obtaining the first sample quality score corresponding to the local sample data features of the prediction request terminal and the second sample quality score corresponding to the local sample data features of each target prediction terminal, the method further includes: Based on multiple preset quality evaluation dimensions, the actual feature measurement value corresponding to each preset quality evaluation dimension is extracted from the local sample data features of each prediction end. For each prediction end, the evaluation score of the actual feature metric corresponding to each preset quality evaluation dimension is determined based on the mapping relationship between the feature metric value and the preset score of the preset quality evaluation dimension. The sample quality score is obtained by summing the evaluation scores of the actual feature measurement values ​​corresponding to each preset quality evaluation dimension. The step of obtaining the first sample quality score corresponding to the local sample data features of the prediction request terminal and the second sample quality score corresponding to the local sample data features of each target prediction terminal includes: The first sample quality score corresponding to the local sample data features of the prediction request terminal and the second sample quality score corresponding to the local sample data features of each target prediction terminal are obtained from multiple sample quality scores.

6. The method as described in claim 1, characterized in that, Sending the target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence level to the prediction request terminal includes: Generate risk interpretation information, which includes several clinical features that have the greatest impact on the probability of the target cerebrovascular disease risk and their contribution; The risk interpretation information, along with the target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence level, are sent to the prediction request terminal.

7. The method as described in claim 1, characterized in that, The process of receiving the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence level returned by each target prediction terminal includes: The response status of each target prediction end is detected; If at least one target prediction end fails to return the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence within the preset response time, the unresponsive target prediction end will be removed from the participation list of the current collaborative prediction task. Receive the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence level returned by the remaining target prediction terminals that have responded.

8. The method as described in claim 1, characterized in that, Sending a second collaborative prediction request to each target prediction endpoint includes: The clinical feature vector is encrypted to obtain an encrypted clinical feature vector; A second collaborative prediction request is generated based on the encrypted clinical feature vector; Send a second collaborative prediction request to each target prediction endpoint.

9. The method as described in claim 1, characterized in that, Sending the target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence level to the prediction request terminal includes: When the overall prediction confidence is less than the preset confidence threshold, an unreliable prompt is generated for the target cerebrovascular disease risk probability, and a data supplement prompt is generated for the clinical feature data. Send the unreliability alert and the data supplementation alert to the prediction request terminal; When the overall prediction confidence level is not less than the preset confidence threshold, the target cerebrovascular disease risk probability and its corresponding overall prediction confidence level are sent to the prediction request terminal.

10. A device for predicting the risk of cerebrovascular disease based on a distributed biobank, characterized in that, The device includes: The request receiving unit is used to receive a first collaborative prediction request sent by the prediction requesting end. The first collaborative prediction request carries a clinical feature vector, a first initial cerebrovascular disease risk probability and its corresponding first prediction confidence. The clinical feature vector is extracted by the prediction requesting end based on the clinical feature data of the object to be diagnosed. The first initial cerebrovascular disease risk probability and its corresponding first prediction confidence are generated by the prediction requesting end based on the clinical feature vector and using its locally deployed cerebrovascular disease risk prediction model. A matching unit is configured to determine at least one target prediction end that matches the clinical feature vector from multiple candidate prediction ends based on the first collaborative prediction request and the local sample data features of multiple candidate prediction ends. The local sample data features are used to characterize the data attributes of the local biobank used by the corresponding candidate prediction end when training its locally deployed cerebrovascular disease risk prediction model. The request sending unit is used to send a second collaborative prediction request to each target prediction end. The second collaborative prediction request carries the clinical feature vector, so that each target prediction end can generate a corresponding second initial cerebrovascular disease risk probability and its corresponding second prediction confidence based on the clinical feature vector and its locally deployed cerebrovascular disease risk prediction model. The data receiving unit is used to receive the second initial cerebrovascular disease risk probability and its corresponding second prediction confidence level returned by each target prediction terminal; The weighted fusion unit is used to fuse the first prediction confidence and each of the second prediction confidence to obtain a comprehensive prediction confidence, and to perform weighted fusion of the first initial cerebrovascular disease risk probability and each of the second initial cerebrovascular disease risk probabilities to obtain a target cerebrovascular disease risk probability. The data sending unit is used to send the target cerebrovascular disease risk probability and its corresponding comprehensive prediction confidence to the prediction request terminal.

Images