Medical data processing method and system and related equipment

Abstract

The invention provides a medical data processing method and system and related equipment, and relates to the technical field of medical data processing. The method comprises the steps that multi-modal medical data of a target object are acquired, and the multi-modal medical data comprise at least one of clinical medical data, medical image data and biomarker data; inputting the multi-modal medical data into a multi-modal feature extraction model, and outputting multi-modal feature data related to intracranial atherosclerotic stenosis (ICAS) stroke risk prediction; and inputting the multi-modal feature data into a pre-trained ICAS stroke risk prediction model, and outputting an ICAS stroke risk prediction result of the target object. According to the invention, deep integration and accurate feature mining of multi-modal medical data can be realized, the efficiency and accuracy of ICAS stroke risk prediction are significantly improved, and decision support is provided for a clinician to make a personalized treatment scheme.

Description

Technical Field

[0001] This disclosure relates to the field of medical data processing technology, and in particular to a medical data processing method, system and related equipment. Background Technology

[0002] Intracranial atherosclerotic stenosis (ICAS) is a significant factor contributing to stroke recurrence and disability. With an aging population, the incidence of ICAS is rising annually, becoming one of the major challenges facing clinical neurologists. Existing stroke risk assessment methods rely on clinical medical data, resulting in limited accuracy.

[0003] CT (Computed Tomography) and MRI (Magnetic Resonance Imaging) are two commonly used imaging techniques, widely applied in the diagnosis and prediction of various diseases due to their ability to extract quantitative features. However, information such as vascular stenosis and cerebral blood flow obtained through CT and MRI is insufficient to comprehensively reflect stroke risk because it cannot be combined with medical clinical data or biomarker data.

[0004] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] This disclosure provides a medical data processing method, system, and related equipment, which at least to some extent overcomes the technical problem of low accuracy in stroke risk assessment methods in related technologies.

[0006] Other features and advantages of this disclosure will become apparent from the following detailed description, or may be learned in part by practice of this disclosure.

[0007] According to one aspect of this disclosure, a medical data processing method is provided, comprising: acquiring multimodal medical data of a target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; inputting the multimodal medical data into a multimodal feature extraction model, and outputting multimodal feature data related to intracranial atherosclerotic stenosis (ICAS) stroke risk prediction; and inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model, and outputting the ICAS stroke risk prediction result of the target object.

[0008] In some embodiments, the multimodal medical data includes: multimodal medical data of the target object in different time periods; wherein, inputting the multimodal medical data into a multimodal feature extraction model and outputting multimodal feature data related to intracranial atherosclerotic stenosis (ICAS) stroke risk prediction includes: inputting the multimodal medical data of the target object in different time periods into the multimodal feature extraction model and outputting the multimodal feature data of the target object in different time periods; fusing the same-modal feature data of the target object in different time periods to obtain the multimodal fusion feature of the target object; inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result of the target object includes: inputting the multimodal fusion feature of the target object into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result of the target object.

[0009] In some embodiments, after inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result for the target object, the method further includes: generating a personalized treatment plan for the target object based on the ICAS stroke risk prediction result for the target object, wherein the personalized treatment plan includes: a treatment plan for the target object in at least one treatment stage; obtaining the treatment feedback result of the target object in the first treatment stage, and updating the treatment plan for the target object in the second treatment stage based on the treatment feedback result of the target object in the first treatment stage, wherein the first treatment stage is the previous treatment stage of the second treatment stage.

[0010] In some embodiments, after inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result for the target object, the method further includes: visually displaying the ICAS stroke risk prediction result for the target object.

[0011] In some embodiments, acquiring multimodal medical data of a target object includes: generating one or more prompts based on a pre-trained medical language model and predetermined ICAS stroke risk prediction indicators, wherein the medical language model is obtained by fine-tuning a large language model using data from the medical field, and the prompts are used to prompt the user to input indicator data corresponding to each reference indicator; and acquiring multimodal medical data input by the user based on each prompt.

[0012] According to another aspect of this disclosure, a medical data processing system is also provided, comprising: a clinical medical data module for providing clinical medical data of a target object; a medical imaging data module for providing medical imaging data of the target object; a biomarker data module for providing biomarker data of the target object; and a data processing module for inputting the data provided by the clinical medical data module, the clinical medical data module, and the biomarker data module into a multimodal feature extraction model, outputting multimodal feature data related to ICAS stroke risk prediction, and inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model to output the ICAS stroke risk prediction result of the target object.

[0013] According to another aspect of this disclosure, a medical data processing apparatus is also provided, comprising: a multimodal medical data acquisition module for acquiring multimodal medical data of a target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; a multimodal feature extraction module for inputting the multimodal medical data into a multimodal feature extraction model and outputting multimodal feature data related to intracranial atherosclerotic stenosis (ICAS) stroke risk prediction; and a prediction module for inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result of the target object.

[0014] According to another aspect of this disclosure, an electronic device is also provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the medical data processing method described in any of the preceding claims by executing the executable instructions.

[0015] According to another aspect of this disclosure, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the medical data processing method described in any of the preceding claims.

[0016] According to another aspect of this disclosure, a computer program product is also provided, comprising: a computer program or instructions that, when executed by a processor, implement the medical data processing method described in any one of the preceding claims.

[0017] The medical data processing method, system, and related equipment provided in this disclosure, by pre-training a multimodal feature extraction model capable of extracting feature data related to ICAS stroke risk prediction from multimodal medical data, and an ICAS stroke risk prediction model capable of predicting ICAS stroke risk based on multimodal feature data, can quickly and accurately extract multimodal feature data and perform ICAS stroke risk prediction after obtaining multimodal medical data such as clinical medical data, medical imaging data, and biomarker data of the target object.

[0018] This disclosure enables deep integration and precise feature mining of multimodal medical data, significantly improving the efficiency and accuracy of ICAS stroke risk prediction and providing decision support for clinicians to develop personalized treatment plans. This disclosure can be applied to, but is not limited to, stroke risk assessment and personalized treatment decision support systems, particularly in clinical research and applications in the fields of neurology, neuroimaging, and bioinformatics.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0021] Figure 1 This diagram illustrates an application system architecture according to an embodiment of the present disclosure. Figure 2 This diagram illustrates a flowchart of a medical data processing method according to an embodiment of the present disclosure. Figure 3 A flowchart of an optional medical data processing method according to an embodiment of this disclosure is shown; Figure 4 A flowchart of an optional medical data processing method according to an embodiment of this disclosure is shown; Figure 5 A flowchart illustrating the extraction of a time-dimensional feature in an embodiment of this disclosure is shown. Figure 6 This diagram illustrates a method for generating a treatment plan according to an embodiment of the present disclosure; Figure 7 This diagram illustrates a method for acquiring multimodal medical data according to an embodiment of the present disclosure. Figure 8This diagram illustrates an interface interaction diagram of acquiring multimodal medical data based on a medical large language model, according to an embodiment of this disclosure. Figure 9 This illustration shows a specific implementation framework for predicting ICAS stroke risk based on clinical medical data, medical imaging data, and biomarker data, according to an embodiment of this disclosure. Figure 10 This diagram illustrates an ICAS stroke risk prediction model architecture according to an embodiment of the present disclosure. Figure 11 This diagram illustrates a data preprocessing and feature extraction flowchart according to an embodiment of the present disclosure; Figure 12 This diagram illustrates a machine learning model training and prediction process according to an embodiment of the present disclosure. Figure 13 This diagram illustrates a stroke risk assessment result according to an embodiment of the present disclosure. Figure 14 A schematic diagram of the interface for generating a personalized treatment plan is shown in an embodiment of this disclosure; Figure 15 This diagram illustrates a medical data processing system according to an embodiment of the present disclosure. Figure 16 This diagram illustrates a medical data processing apparatus according to an embodiment of the present disclosure. Figure 17 A schematic diagram of an electronic device according to an embodiment of the present disclosure is shown. Detailed Implementation

[0022] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0023] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0024] To facilitate understanding, before introducing the embodiments of this disclosure, the following explanations are provided for several terms involved in the embodiments of this disclosure: ICAS: Intracranial Atherosclerotic Stenosis.

[0025] CT: Computed Tomography.

[0026] MRI: Magnetic Resonance Imaging.

[0027] ICAS (Intravascular Coagulation) is one of the leading causes of stroke recurrence and disability. Traditional stroke risk assessment methods often rely on single imaging examinations or clinical data, failing to comprehensively reflect a patient's stroke risk and lacking precise assessment tailored to individual patients. Existing imaging techniques such as CT and MRI can provide information on the degree of vascular stenosis and cerebral blood flow, but they cannot effectively combine biomarkers and clinical data, resulting in a lack of personalization and accuracy in stroke risk assessment. Furthermore, although some machine learning algorithms have been applied to medical imaging data analysis, these techniques mostly remain at the level of single data source applications, lacking sufficient multimodal data fusion and large-scale multicenter data support. Therefore, how to effectively integrate imaging data, clinical data, and biomarker data to provide a more accurate and personalized stroke risk prediction model remains a pressing issue that needs to be addressed.

[0028] This disclosure proposes an ICAS stroke risk prediction model based on multimodal data. It combines radiomics, clinical data, and biomarkers, and uses machine learning algorithms to perform multidimensional analysis on these data to construct a personalized stroke risk assessment system.

[0029] The specific implementation methods of the embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0030] Figure 1 A schematic diagram of an exemplary application system architecture to which the medical data processing methods of the embodiments of this disclosure can be applied is shown. Figure 1 As shown, the system architecture may include a terminal device 10 and a server 20.

[0031] In this embodiment of the disclosure, the medium providing the communication link between the terminal device 10 and the server 20 can be a wired network or a wireless network.

[0032] Optionally, the aforementioned wireless or wired networks use standard communication technologies and / or protocols. The network is typically the Internet, but can also be any network, including but not limited to Local Area Networks (LANs), Metropolitan Area Networks (MANs), Wide Area Networks (WANs), mobile, wired or wireless networks, private networks, or any combination of virtual private networks. In some embodiments, technologies and / or formats, including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network. Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Networks (VPNs), and Internet Protocol Security (IPSec) can be used to encrypt all or some links. In other embodiments, custom and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0033] Terminal device 10 can be various electronic devices, including but not limited to smartphones, tablets, laptops, desktop computers, smart speakers, smartwatches, wearable devices, augmented reality devices, virtual reality devices, etc.

[0034] Optionally, the client of the application installed on different terminal devices 10 may be the same, or the client of the same type of application based on different operating systems. Depending on the terminal platform, the specific form of the application client may also be different; for example, the application client may be a mobile client, a PC client, etc.

[0035] Server 20 can be a server that provides various services, such as a backend management server that supports the device operated by the user using terminal device 10. The backend management server can analyze and process received requests and other data, and feed the processing results back to the terminal device.

[0036] Optionally, the server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0037] Those skilled in the art will know that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative; any number of terminal devices, networks, and servers can be included depending on actual needs. This disclosure does not limit the scope of the embodiments.

[0038] Under the above system architecture, this disclosure provides a medical data processing method that can be executed by any electronic device with computing capabilities.

[0039] In some embodiments, the medical data processing method provided in this disclosure can be executed by a terminal device of the system architecture described above; in other embodiments, the medical data processing method provided in this disclosure can be executed by a server in the system architecture described above; in still other embodiments, the medical data processing method provided in this disclosure can be implemented by the terminal device and the server in the system architecture described above through interaction.

[0040] This disclosure, for the first time, combines radiomics, clinical data, and biomarker data, employing machine learning techniques to construct an ICAS stroke risk prediction model based on multimodal medical data. By combining radiomics, clinical data, and biomarkers, and using machine learning algorithms to perform multidimensional analysis of these data, a personalized stroke risk prediction model is built. By integrating patient imaging characteristics, clinical information, and biomarker data, the model can provide personalized stroke risk predictions, helping physicians develop optimal treatment plans for each patient. This model provides clinicians with a precise decision support tool, helping them develop more personalized treatment plans, contributing to reducing stroke incidence and improving patients' quality of life and prognosis. This disclosure can be applied to, but is not limited to, ICAS patient groups in different regions and among different populations.

[0041] Figure 2 This invention discloses a flowchart of a medical data processing method according to an embodiment of the present invention, as shown below. Figure 2 As shown, the medical data processing method provided in this embodiment includes the following steps: S202, acquire multimodal medical data of the target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data.

[0042] It should be noted that the target group can be any individual from any population group, including healthy individuals and patients. In this embodiment, the extracted medical imaging data features include, but are not limited to, vascular stenosis, hemodynamic characteristics, and imaging features of the lesion area. The biomarker data mentioned above can be data obtained from the analysis of blood samples.

[0043] S204 inputs multimodal medical data into a multimodal feature extraction model and outputs multimodal feature data related to the prediction of stroke risk in intracranial atherosclerotic stenosis (ICAS).

[0044] In practice, a single multimodal feature extraction model can be used to adaptively identify and extract features from multimodal medical data, or multiple feature extraction models can be used to extract features from different modalities of medical data. Optionally, the feature extraction model can be obtained by training a convolutional neural network.

[0045] In some embodiments, when using a single model to adaptively identify multimodal medical data, a multimodal feature extraction master model can be constructed. This model is trained on a convolutional neural network (CNN) architecture. By adding cross-modal attention layers and feature fusion layers, and configuring model parameters (including but not limited to: convolutional kernel size, number of hidden layer neurons, and attention weight calculation threshold), the multimodal feature extraction model can adaptively identify the type differences of medical image data, clinical medical data, and biomarker data and extract the corresponding feature data respectively.

[0046] In other embodiments, when multiple feature extraction models are used to extract features from medical data of different modalities, three independent feature extraction sub-models can be constructed: a medical image feature extraction sub-model, a clinical medicine feature extraction sub-model, and a biomarker feature extraction sub-model. The medical image feature extraction sub-model is trained on a CNN (such as ResNet or U-Net architecture) model to extract morphological and structural features of intracranial artery images. The clinical medicine feature extraction sub-model is trained on a model combining CNN and fully connected layers to extract structured and correlated features of clinical medicine data. The biomarker feature extraction sub-model is trained on a lightweight CNN to extract numerical fluctuations and correlation features of biomarker data.

[0047] It should be noted that, if an adaptive single-model implementation is adopted, the main multimodal feature extraction model adaptively matches the feature extraction logic of each modality of data to output multimodal medical feature data; if a multi-model collaborative implementation is adopted, three feature extraction sub-models respectively extract features from the data of the corresponding modality to obtain medical image feature data, clinical medical feature data, and biomarker feature data. Then, the medical image feature data, clinical medical feature data, and biomarker feature data are aligned using a feature alignment algorithm (such as cosine similarity alignment) to form multimodal medical feature data.

[0048] The adaptive single-model implementation achieves full-modal data processing through a single multimodal feature extraction master model, eliminating the need to build multiple independent sub-models and complex cross-model communication mechanisms. The master model is based on a trained CNN, requiring only the addition of cross-modal attention layers and feature fusion layers to adapt to three types of data: imaging, clinical, and biomarkers. This reduces the number of models and the corresponding parameter management costs, lowers the hardware resource requirements during system deployment (such as server computing power allocation and storage space requirements), and simplifies subsequent maintenance by requiring only version updates and troubleshooting for a single model. Compared to multiple models that require subsequent integration through feature alignment algorithms, a single model can simultaneously complete cross-modal association learning during feature extraction, improving the timeliness and accuracy of feature fusion.

[0049] The multi-model collaborative implementation approach allows each sub-model to focus on a single modality of data, avoiding generalization or accuracy issues that can arise from a single model needing to accommodate multiple modalities. When the clinical data standards or feature requirements for a particular modality change, only the corresponding sub-model needs to be optimized, without adjusting the entire model system. If a sub-model malfunctions (e.g., the biomarker sub-model is interrupted due to abnormal data format), only the feature extraction for that modality is affected. The feature root extraction sub-models for medical imaging and clinical medicine can still function normally and output feature data, avoiding the risk of a single model failure causing the entire system to crash.

[0050] S206 inputs multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputs the ICAS stroke risk prediction results for the target object.

[0051] In practice, the above can be obtained by training any of the following models: Support Vector Machine (SVM), Random Forest, Gradient Boosting Tree (XGBoost), Convolutional Neural Network (CNN), or Long Short-Term Memory Network (LSTM).

[0052] Optionally, in some embodiments, the model can also be updated online based on new patient data through an incremental learning mechanism. By introducing a follow-up data feedback mechanism and online learning algorithms, continuous iterative optimization of the model can be achieved, effectively improving its long-term adaptability and predictive reliability in real clinical environments, and avoiding the problem of static models becoming detached from real-world scenarios.

[0053] As can be seen from the above, the medical data processing method provided in this embodiment collects imaging data, clinical data, and biomarker data, and uses machine learning algorithms to train the data to generate a stroke risk prediction model. The imaging data includes imaging information such as CT, MRI, and angiography; the biomarker data includes blood sample analysis results; and the clinical data includes the patient's medical history and clinical manifestations. Key imaging features are extracted through radiomics, and the biomarker data and clinical data are processed through feature selection to obtain variables related to stroke risk. These variables are then combined with the machine learning model for risk assessment, generating a personalized stroke risk score. This model provides stroke risk prediction for ICAS patients and provides decision support for clinicians in developing personalized treatment plans.

[0054] The embodiments disclosed herein propose innovative solutions in several key aspects such as data fusion, model building, interface interaction, and model optimization. These solutions not only improve the technical level of ICAS stroke risk prediction but also provide a feasible example for the clinical translation of artificial intelligence in the field of precision neurovascular medicine.

[0055] In some embodiments, such as Figure 3 As shown, after inputting multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result for the target object, the medical data processing method provided in this embodiment may further include the following steps: S208 visualizes the ICAS stroke risk prediction results for the target object.

[0056] In some embodiments, multimodal medical data includes: multimodal medical data of the target object over different time periods; then, as follows: Figure 4 As shown, the medical data processing method provided in this embodiment can achieve ICAS stroke risk prediction through the following steps: S402, acquire multimodal medical data in different time periods, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; S404: Input the multimodal medical data of the target object in different time periods into the multimodal feature extraction model, and output the multimodal feature data of the target object in different time periods; S406, fuse the same-modal feature data of the target object in different time periods to obtain the multimodal fusion feature of the target object; S408 inputs the multimodal fusion features of the target object into the pre-trained ICAS stroke risk prediction model and outputs the ICAS stroke risk prediction result of the target object.

[0057] In the above embodiments, for data of the same modality, data acquired at different time periods are fused in the time dimension. This considers not only horizontal multimodal fusion but also vertical time dimension fusion. This combination of time dimension and different modal dimensions enables more accurate prediction results. In real-world scenarios, patients often undergo regular physical examinations or checkups. It may be difficult to detect lesion features based on a single CT or MRI image, but it is much easier to detect lesion features based on multiple CT or MRI images. Figure 5 As shown, for a specific target object, its multimodal medical data across N time periods can be fused along the time dimension to extract fused features. It should be noted that the fusion along the time dimension can be a weighted result of features from multiple time periods or an incremental result; this disclosure does not impose specific limitations on this.

[0058] Optionally, when fusing multimodal medical feature data across different time periods, different weights can be configured based on the distance between the time period corresponding to the multimodal medical feature data and the time of occurrence of the abnormality (such as lesions). For example, if an abnormality is detected in the 6th time period among 10 time periods for a target object, when fusing the multimodal medical feature data of these 10 time periods for the target object, the weights corresponding to the 1st, 2nd, 3rd, 4th, and 5th time periods can be set to increase sequentially; the weights corresponding to the 7th, 8th, and 9th time periods can be set to decrease sequentially; and the weight corresponding to the 6th time period can be the largest.

[0059] Since different modalities of medical data correspond to different feature data, in some embodiments, the same modal feature data of the target object in different time periods are fused to obtain the multimodal fused features of the target object. When fusing multimodal medical data in different time periods in the time dimension, it can be achieved through the following steps: For clinical medical feature data extracted from clinical medical data, an Attention-Based Long Short-Term Memory Network (Attention-LSTM) model is used for temporal fusion. By allocating attention weights, the contribution of key clinical medical features within different time periods is highlighted, resulting in fused clinical medical feature data. When processing clinical medical feature data, the Attention-LSTM model automatically assigns weights to clinical medical features across different time periods using its attention mechanism. For example, when fusing symptom descriptions and vital signs data from different time periods, it can focus on features highly correlated with the patient's condition (such as temperature and blood pressure changes during acute illness) while reducing interference from irrelevant or secondary features (such as routine physical examination data without abnormalities). This makes the fusion results more aligned with medical analysis needs and provides more targeted feature support for subsequent disease diagnosis.

[0060] For medical image feature data extracted from medical image data, a Temporal Convolutional Network (TCN) model is used for temporal fusion. The TCN model utilizes convolutional kernels to capture the temporal dependencies of medical image features across different time periods, outputting fused medical image feature data. Through the sliding operation of the convolutional kernels, TCN can capture both local temporal dependencies of medical image features in adjacent time periods (such as subtle morphological changes in lesions within a short period) and global temporal dependencies of image features over longer time spans (such as the overall growth trend of tumors over several months) by expanding the receptive field through convolution. This characteristic aligns perfectly with the need for medical image data (such as CT and MRI images) to simultaneously focus on local details and overall changes, enabling a more comprehensive fusion of image features from different time periods. The convolutional operations of the TCN model can be parallelized, significantly reducing data processing time and improving overall fusion efficiency when processing large-scale medical image data (such as multi-time period image data of batches of patients), meeting the speed requirements for data processing in medical scenarios.

[0061] For biomarker feature data extracted from biomarker data, a Gated Recurrent Unit Model (GRU) is used for time fusion. The effective information of biomarker features in different time periods is filtered through the gating mechanism, and the fused biomarker feature data is output.

[0062] Optionally, in some embodiments, cross-validation is used to optimize the parameters of the Attention-LSTM model, TCN model, and GRU model. By adjusting the number of hidden layer neurons, learning rate, and number of iterations, the accuracy of the fusion results of each model meets a preset threshold. The Attention-LSTM model, TCN model, and GRU model used all have efficient temporal data processing capabilities. Optimization of model parameters can further improve computational efficiency and meet the batch processing needs of large-scale multimodal medical data. At the same time, the system adopts a parallel processing architecture, allowing data acquisition, feature extraction, and temporal fusion to proceed simultaneously, reducing data processing delays and ensuring timely output of fusion results, providing rapid data support for emergency medical scenarios (such as emergency condition assessment).

[0063] In this embodiment, suitable temporal fusion models (Attention-LSTM model, TCN model, and GRU model) are adopted for different types of multimodal medical feature data. These models can accurately capture the temporal patterns and key information of each feature data. Combined with model optimization steps, the impact of unreasonable model parameters on the fusion results can be effectively reduced, making the output fused feature data more consistent with the actual physiological and pathological state of the target object, providing high-quality data support for subsequent medical analysis. The multimodal feature data after temporal fusion can fully reflect the trend of health status changes of the target object in different time periods. For example, by fusing medical image features from different time periods, the development process of lesions can be clearly presented, and by fusing biomarker features, the dynamic changes of the disease can be accurately reflected. Compared with the LSTM model, the GRU model simplifies the gating structure (containing only update and reset gates), reducing the number of model parameters and computational complexity while ensuring effective processing of biomarker feature data. For biomarker data (such as protein marker concentration and nucleic acid marker activity data in blood tests), which require frequent collection and processing, temporal fusion can be completed with lower computational resource consumption, reducing system operating costs. The gating mechanism can flexibly filter effective information on biomarker characteristics at different time periods and eliminate abnormal fluctuations or irrelevant data (such as abnormal values ​​of biomarker concentration caused by detection errors).

[0064] In some embodiments, such as Figure 6 As shown, after outputting the ICAS stroke risk prediction results for the target object, the medical data processing method provided in this embodiment may further include the following steps: S602, Based on the ICAS stroke risk prediction results of the target object, generate a personalized treatment plan for the target object, wherein the personalized treatment plan includes: the treatment plan for the target object in at least one treatment stage; S604, obtain the treatment feedback results of the target object in the first treatment stage, and update the treatment plan of the target object in the second treatment stage according to the treatment feedback results of the target object in the first treatment stage, wherein the first treatment stage is the previous treatment stage of the second treatment stage.

[0065] For ICAS stroke patients, multiple treatment phases are often required. Therefore, after obtaining the ICAS stroke risk prediction results for the target individual, a personalized treatment plan can be generated for that individual. If the generated personalized treatment plan includes multiple treatment phases, the treatment plan for the next treatment phase can be updated based on the treatment feedback after a certain treatment phase is implemented, thereby achieving dynamic updating of the personalized treatment plan.

[0066] In some embodiments, such as Figure 7 As shown in the embodiments of this disclosure, the medical data processing method can acquire multimodal medical data of a target object through the following steps: S702, based on a pre-trained medical big language model, generates one or more prompts according to pre-determined ICAS stroke risk prediction indicators. The medical big language model is obtained by fine-tuning the big language model using data from the medical field, and the prompts are used to prompt the user to input the indicator data corresponding to each reference indicator. S704, acquire multimodal medical data input by the user based on various prompts.

[0067] It should be noted that for a specific target group, there may be a large amount of multimodal medical data, but not all of this multimodal medical data is related to ICAS stroke. Therefore, how to quickly obtain ICAS stroke-related data from the patient's multimodal medical data is crucial for achieving rapid and accurate ICAS stroke prediction. In this embodiment, the ability of a large language model to quickly process data is leveraged. The large language model is fine-tuned using data from the medical field to obtain a medical large language model. This medical large language model is then used to allow users to input ICAS stroke-related data through a question-and-answer format.

[0068] In some embodiments, users may input data in any of the following ways, but not limited to: ① replying with content, including text content and / or voice content; ② uploading file data, including one or more of documents, videos and images; ③ acquiring data by taking a picture or scanning.

[0069] As can be seen, prompting users to input data through a question-and-answer format requires very little expertise from the user. Users only need to upload the relevant data according to the question prompts, and the backend system will automatically identify and analyze the uploaded data. If the uploaded data lacks relevant information, the system can ask the user further questions until it obtains the data related to ICAS stroke risk prediction.

[0070] Figure 8 This illustration shows a schematic diagram of an interface interaction for acquiring multimodal medical data based on a large medical language model, as shown in an embodiment of this disclosure. Figure 8 As shown, users can input data through various methods such as text input, voice input, importing from the album, direct shooting, and uploading files.

[0071] It should be noted that, Figure 8 The dialog interface shown is merely an example, and those skilled in the art can adjust it according to the actual situation. This disclosure does not impose any specific limitations on it.

[0072] To avoid the leakage of sensitive data (such as user privacy data), in some embodiments, when obtaining multimodal medical data input by the user based on various prompts, privacy desensitization processing can be performed on the multimodal medical data input by the user. For example, the entity replacement method can be used to replace privacy entities such as patient name, contact information, and medical institution in the multimodal medical data input by the user with standardized placeholders (such as "[patient name]" "[medical institution]").

[0073] Figure 9 This illustration shows a schematic diagram of a specific implementation framework for ICAS stroke risk prediction based on clinical medical data, medical imaging data, and biomarker data, as shown in this embodiment of the present disclosure. Figure 9 As shown in the embodiments of this disclosure, a stroke risk prediction model and assessment method for intracranial atherosclerotic stenosis (ICAS) based on multimodal imaging and biomarkers are provided. This method collects multi-center clinical data, including imaging data (such as CT and MRI), biomarker data (such as blood sample analysis), and clinical data, and combines this with machine learning algorithms to establish a stroke risk prediction model. The model extracts hemodynamic characteristics of vascular stenosis and lesion areas, as well as biomarkers related to stroke risk, using radiomics technology to comprehensively analyze the patient's stroke risk. By evaluating the effects of different treatment methods (such as drug therapy, endovascular therapy, and surgical treatment), personalized treatment strategies are provided to optimize clinical decision-making. Through these embodiments, accurate stroke risk assessment and personalized treatment plans can be provided for ICAS patients. It is mainly applied to stroke risk prediction and the development of personalized treatment strategies, and is suitable for multi-center medical research and hospital clinical practice.

[0074] The specific implementation steps are as follows: 1) Imaging data processing: Patients are examined using CT or MRI imaging techniques to extract quantitative features related to vascular stenosis and cerebral blood flow dynamics. Radiomics methods are used to extract high-dimensional features from the imaging data, such as vascular morphology, blood flow velocity, and cerebral perfusion.

[0075] 2) Integration of clinical data and biomarkers: Collect patients' clinical data, including past medical history, clinical symptoms, family history, etc., and analyze relevant biomarkers (such as inflammatory factors, coagulation factors, etc.) from blood samples. Select variables closely related to stroke risk.

[0076] 3) Data Fusion and Machine Learning Model Building: Imaging features, clinical data, and biomarker data are fused together, and machine learning algorithms (such as support vector machines, random forests, or deep neural networks) are used to train the data to build a comprehensive and personalized stroke risk prediction model. This model can automatically identify and quantify important information from different data sources and comprehensively assess a patient's stroke risk.

[0077] 4) Predictive results and clinical decision support: Based on the trained predictive model, a stroke risk score is generated for each patient, and personalized clinical decision support is provided to doctors to help develop treatment strategies.

[0078] In this embodiment, a predictive model for assessing ICAS stroke risk is constructed based on multi-center cohort data and machine learning algorithms. This model integrates imaging data, clinical data, and biomarker data to provide accurate stroke risk prediction for clinicians and assist in developing personalized treatment plans. Figure 10 This diagram illustrates an ICAS stroke risk prediction model architecture according to an embodiment of the present disclosure. Figure 11 This diagram illustrates a data preprocessing and feature extraction flowchart according to an embodiment of the present disclosure. Figure 12 This diagram illustrates a machine learning model training and prediction process according to an embodiment of the present disclosure. Figure 13 A schematic diagram of a stroke risk assessment result in an embodiment of this disclosure; Figure 14 A schematic diagram of the interface for generating a personalized treatment plan is shown in an embodiment of this disclosure.

[0079] In this embodiment, radiomics technology is used to automatically extract cerebral vascular stenosis, blood flow perfusion and structural features. At the same time, deep neural networks (such as U-Net and ResNet) are introduced to assist in the segmentation of key regions and parameter extraction, which significantly reduces the annotation burden on doctors and improves processing consistency and efficiency.

[0080] The following is combined Figures 10-14 This will provide a detailed introduction to the specific implementation plan, including: 1) Data Acquisition and Preprocessing: ① Data Acquisition: Data sources include, but are not limited to: imaging data (such as MRI and CT images), clinical data (such as medical records, diagnostic records, and medication history), and biomarker data (such as blood chemistry indicators). All data undergoes standardization to ensure that data from different centers can be integrated and analyzed on the same platform. After data acquisition, the following operations need to be performed, but are not limited to: ② Data preprocessing: Image data standardization: Each image dataset undergoes preprocessing, including denoising, spatial alignment, and standardization, to ensure data comparability.

[0081] Image data feature extraction: Quantitative features of cerebral blood vessels and brain tissue are extracted using radiomics algorithms, such as the size of atherosclerotic plaques, the degree of vascular stenosis, arterial wall thickness, and cerebral hemodynamic characteristics. These features are crucial for predicting stroke risk.

[0082] Deep learning feature extraction: For complex patterns in image data, convolutional neural networks (CNNs) can be used for automatic feature extraction to improve the efficiency and accuracy of feature extraction.

[0083] Clinical data and biomarker processing: Clinical data includes common stroke risk factors such as age, sex, smoking history, hypertension, and diabetes. This data was transformed into structured data after removing missing values ​​and standardization. Biomarker data (such as blood lipids, blood glucose, and blood pressure) were converted into numerical features and combined with imaging and clinical data for multimodal analysis.

[0084] Data fusion: This involves fusing imaging data, clinical data, and biomarker data into a single high-dimensional feature vector, which serves as input to a machine learning model. This fusion process utilizes feature selection algorithms to remove redundant features, ensuring the model's simplicity and accuracy.

[0085] 2) Training and optimization of machine learning models: ① Model Selection and Training: After data processing, this invention uses several common machine learning algorithms for training, including Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting Decision Tree (XGBoost). The specific steps are as follows: SVM: Used to process high-dimensional data. It maps data to a high-dimensional space through kernel functions to achieve non-linear classification.

[0086] Random Forest: It utilizes the idea of ​​ensemble learning to improve the robustness and accuracy of the model through a voting mechanism of multiple decision trees.

[0087] XGBoost: As an efficient ensemble algorithm, it generates the final prediction result by weighted combination of weak classifiers, and is particularly suitable for large-scale datasets.

[0088] ② Model Training: Up to 80% of the data is used for training, with the remaining 20% ​​used for validation and testing. Each model is evaluated using k-fold cross-validation. Hyperparameters are tuned using grid search and random search to ensure optimal performance on different datasets.

[0089] ③ Model evaluation and optimization: Evaluation metrics: Accuracy, precision, recall, F1 score, and AUC (area under the curve) are used to comprehensively evaluate the model's performance.

[0090] Optimization strategy: To address the overfitting problem, use regularization techniques (such as L2 regularization) to control model complexity; at the same time, further optimize the model through feature selection and dimensionality reduction methods to reduce computational burden and improve the model's generalization ability.

[0091] 3) Risk prediction and clinical decision support: ① Stroke Risk Prediction: This embodiment uses a trained model to assess the risk of new ICAS patient data. The input data undergoes the same preprocessing and feature extraction process as the training data before being fed into the model for prediction. The model output is a stroke risk score, ranging from 0 to 1, where 0 represents the lowest risk and 1 represents the highest risk. Based on the score, patients are categorized into low-risk, intermediate-risk, and high-risk groups.

[0092] ② Clinical Decision Support: Based on the patient's risk score, the system automatically provides personalized recommendations to clinicians. For example, low-risk patients can choose routine follow-up; high-risk patients are advised to undergo further revascularization therapy or intensified antiplatelet therapy. The system can also generate treatment recommendations based on the patient's specific circumstances (such as age, medical history, etc.) and automatically track the patient's subsequent treatment effects.

[0093] ③ Data Visualization: Provides a user-friendly graphical interface that displays stroke risk prediction results, clinical recommendations, and other information to doctors in chart form. For example... Figure 13 As shown, different colors can be displayed for patients with different risk levels, so that doctors can quickly determine the severity of the patient's condition based on the different colors.

[0094] 4) Multi-center data integration and model optimization: ① Data Integration: This embodiment further optimizes the robustness of the model by integrating data from different centers. Since the data sources from different centers vary, standardization and de-identification techniques were employed during the data integration process to ensure patient privacy is protected.

[0095] ② Online Learning and Dynamic Optimization: The model provided in this embodiment supports online learning, allowing for continuous training and optimization as clinical data accumulates. Whenever a new case dataset is added, the model is automatically updated and re-evaluated. Furthermore, an incremental learning strategy is employed to progressively improve the model's predictive capabilities, adapting it to the ever-changing clinical environment.

[0096] ③ Real-time prediction: The patient's stroke risk prediction results can be generated in real time by the system and updated synchronously to the patient's electronic medical record system for use by clinicians.

[0097] After training a stroke risk prediction model, multimodal medical data of patients is acquired, including imaging data (CT, MRI, angiography), biomarker detection data, and clinical information data. Quantitative features are extracted from these data, with imaging data processed using radiomics algorithms and biomarker and clinical data processed using feature engineering. Multiple features are fused into a unified input vector, and feature filtering and dimensionality compression are performed. The trained stroke risk prediction model is then used to output a stroke risk prediction score. Finally, a structured assessment report is generated based on the risk score, and personalized intervention measures are recommended. Figure 14 As shown, it can provide doctors with an interactive interface, allowing them to access data, select models, and view intuitive risk visualization results with a single click. It can also automatically generate structured diagnostic suggestions and personalized intervention plans, lowering the barrier to AI use in clinical settings and increasing the acceptance and adoption rate among medical staff.

[0098] Based on the medical data processing method provided in this embodiment, a platform for assisting doctors in diagnosis and treatment can be provided. Doctors can select patients, automatically retrieve their imaging, experimental, and clinical data, automatically recommend the optimal model and perform predictions, output stroke risk levels and major influencing factors, visualize the assessment results, and generate personalized intervention suggestions. In specific implementations, it can also support the input of actual intervention results and follow-up information for model feedback learning.

[0099] Through the embodiments disclosed herein, the following effects can be achieved, but are not limited to: ① For the first time, deep fusion modeling of multimodal data (images, biomarkers, and clinical information) in ICAS stroke risk prediction is realized. This invention unifies and standardizes the processing of three types of medical data through a customized preprocessing strategy, and constructs a fusion prediction model based on machine learning algorithms. This overcomes the limitations of existing methods that rely solely on single-modal information (such as images or rating scales) for assessment, significantly improving the accuracy and comprehensiveness of risk assessment; ② An automatic feature extraction mechanism combining radiomics and deep learning is introduced to reduce manual intervention and improve processing efficiency. This invention uses radiomics technology to automatically extract cerebral vascular stenosis, blood perfusion, and structural features, while introducing deep neural networks (such as U-Net and ResNet) to assist in key region segmentation and parameter extraction. The system significantly reduces the annotation burden on doctors, improving processing consistency and efficiency; ③ It innovatively designs a visual, user-friendly, interactive intelligent interface to promote the clinical usability of AI models, proposing a risk prediction interface. Doctors can easily access the model and obtain graphical risk assessment results and personalized intervention suggestions with simple operations, breaking down the "black box barrier" of AI in medical applications and enhancing its interpretability and acceptability; ④ It establishes a stroke prediction closed-loop mechanism to achieve dynamic model optimization and individual tracking, supports automatic entry of patient follow-up information, and enables the model to self-update through online learning strategies, solving the problem of "one-time training and long-term fixation" in traditional AI models and improving the continuity and adaptability of predictive performance in real clinical environments; ⑤ It has high scalability, supporting multi-center deployment, federated learning, and mobile terminal usage scenarios. The system can be deployed in hospital information systems, telemedicine platforms, and vehicle / portable devices, supporting multi-center heterogeneous data integration and privacy-preserving modeling, and has broad promotion and transformation value.

[0100] This disclosed embodiment boasts advantages such as high prediction accuracy, user-friendly experience, and sustainable model learning. It can provide ICAS patients with comprehensive and personalized stroke risk assessments and clinical intervention recommendations, and is applicable to various scenarios including hospital information system integration, regional collaborative medical platforms, and remote assisted diagnosis in battlefields or primary care settings, demonstrating broad prospects for widespread application. In specific implementations, a federated learning interface can be reserved to expand the sample size and enhance the model's generalization ability while ensuring patient privacy.

[0101] It should be noted that the acquisition, storage, use, and processing of data in this disclosed technical solution comply with the relevant provisions of national laws and regulations. The various types of data, such as personal identity data, operational data, and behavioral data related to individuals, customers, and groups, obtained in the embodiments of this disclosure have all been authorized.

[0102] Based on the same inventive concept, this disclosure also provides a medical data processing system, as described in the following embodiments. Since the principle by which this system solves the problem is similar to that of the method embodiments described above, the implementation of this system embodiment can refer to the implementation of the method embodiments described above, and repeated details will not be repeated.

[0103] Figure 15 This illustration shows a schematic diagram of a medical data processing system according to an embodiment of the present disclosure, such as... Figure 15 As shown, the system includes: a clinical medical data module 151, a medical imaging data module 152, a biomarker data module 153, and a data processing module 154.

[0104] The system includes a clinical medical data module 151, which provides clinical medical data for the target subject; a medical imaging data module 152, which provides medical imaging data for the target subject; a biomarker data module 153, which provides biomarker data for the target subject; and a data processing module 154, which inputs the data provided by the clinical medical data module, the clinical medical data module, and the biomarker data module into a multimodal feature extraction model, outputs multimodal feature data related to ICAS stroke risk prediction, and inputs the multimodal feature data into a pre-trained ICAS stroke risk prediction model to output the ICAS stroke risk prediction result for the target subject.

[0105] In some embodiments, the multimodal medical data includes: multimodal medical data of the target object in different time periods; then the data processing module 154 is further configured to: input the multimodal medical data of the target object in different time periods into a multimodal feature extraction model, and output the multimodal feature data of the target object in different time periods; fuse the same modal feature data of the target object in different time periods to obtain the multimodal fusion feature of the target object; input the multimodal fusion feature of the target object into a pre-trained ICAS stroke risk prediction model, and output the ICAS stroke risk prediction result of the target object.

[0106] In some embodiments, the data processing module 154 is further configured to: generate a personalized treatment plan for the target object based on the ICAS stroke risk prediction results of the target object, wherein the personalized treatment plan includes: a treatment plan for the target object in at least one treatment stage; obtain the treatment feedback results of the target object in the first treatment stage, and update the treatment plan for the target object in the second treatment stage based on the treatment feedback results of the target object in the first treatment stage, wherein the first treatment stage is the previous treatment stage of the second treatment stage.

[0107] In some embodiments, the data processing module 154 is further configured to: visualize and display the ICAS stroke risk prediction results of the target object.

[0108] In some embodiments, the data processing module 154 is further configured to: generate one or more prompts based on a pre-trained medical big language model and predetermined ICAS stroke risk prediction indicators, wherein the medical big language model is obtained by fine-tuning the big language model using data from the medical field, and the prompts are used to prompt the user to input the indicator data corresponding to each reference indicator; and acquire multimodal medical data input by the user based on each prompt.

[0109] Based on the same inventive concept, this disclosure also provides a medical data processing device, as described in the following embodiments. Since the principle by which this device solves the problem is similar to that of the method embodiments described above, the implementation of this device embodiment can refer to the implementation of the method embodiments described above, and repeated details will not be repeated.

[0110] Figure 16 This illustration shows a schematic diagram of a medical data processing device according to an embodiment of the present disclosure, such as... Figure 16 As shown, the device includes: a multimodal medical data acquisition module 161, a multimodal feature extraction module 162, and a prediction module 163.

[0111] The multimodal medical data acquisition module 161 is used to acquire multimodal medical data of the target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; the multimodal feature extraction module 162 is used to input the multimodal medical data into the multimodal feature extraction model and output multimodal feature data related to the prediction of intracranial atherosclerotic stenosis (ICAS) stroke risk; the prediction module 163 is used to input the multimodal feature data into the pre-trained ICAS stroke risk prediction model and output the ICAS stroke risk prediction result of the target object.

[0112] In some embodiments, the aforementioned multimodal medical data may include: multimodal medical data of the target object in different time periods; then the aforementioned multimodal feature extraction module 162 specifically includes: a feature extraction module 1621, used to input the multimodal medical data of the target object in different time periods into a multimodal feature extraction model, and output multimodal feature data of the target object in different time periods; a feature fusion module 1622, used to fuse the same modal feature data of the target object in different time periods to obtain the multimodal fused features of the target object. In this embodiment, the aforementioned prediction module 163 is further used to: input the multimodal fused features of the target object into a pre-trained ICAS stroke risk prediction model, and output the ICAS stroke risk prediction result of the target object.

[0113] In some embodiments, the medical data processing apparatus provided in this disclosure further includes: a treatment plan generation module 164, configured to: generate a personalized treatment plan for the target object based on the ICAS stroke risk prediction results of the target object, wherein the personalized treatment plan includes: a treatment plan for the target object in at least one treatment stage; obtain the treatment feedback results of the target object in the first treatment stage, and update the treatment plan for the target object in the second treatment stage based on the treatment feedback results of the target object in the first treatment stage, wherein the first treatment stage is the previous treatment stage of the second treatment stage.

[0114] In some embodiments, the medical data processing apparatus provided in this disclosure further includes: a visualization module 165, used to visualize the ICAS stroke risk prediction results of the target object.

[0115] In some embodiments, the multimodal medical data acquisition module 161 is further configured to: generate one or more prompts based on a pre-trained medical language model and predetermined ICAS stroke risk prediction indicators, wherein the medical language model is obtained by fine-tuning the language model using data from the medical field, and the prompts are used to prompt the user to input the indicator data corresponding to each reference indicator; and acquire the multimodal medical data input by the user based on each prompt.

[0116] It should be noted that the examples and application scenarios implemented by the modules in the above device embodiments and the corresponding steps in the method embodiments are the same, but are not limited to the content disclosed in the above method embodiments. It should also be noted that the above modules, as part of the device, can be executed in a computer system such as a set of computer-executable instructions.

[0117] Those skilled in the art will understand that various aspects of this disclosure can be implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which can be collectively referred to herein as a "circuit", "module" or "system".

[0118] Based on the same inventive concept, this disclosure also provides an electronic device, which includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the medical data processing method described above by executing the executable instructions. Since the principle by which this electronic device solves the problem is similar to that of the above method embodiments, the implementation of this electronic device embodiment can refer to the implementation of the above method embodiments, and repeated details will not be described again.

[0119] The following reference Figure 17 To describe an electronic device 1700 according to such an embodiment of the present disclosure. Figure 17 The electronic device 1700 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0120] like Figure 17 As shown, the electronic device 1700 is manifested in the form of a general-purpose computing device. The components of the electronic device 1700 may include, but are not limited to: at least one processing unit 1710, at least one storage unit 1720, and a bus 1730 connecting different system components (including storage unit 1720 and processing unit 1710).

[0121] The storage unit stores program code that can be executed by the processing unit 1710, causing the processing unit 1710 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. For example, the processing unit 1710 can perform the following steps of the above method embodiments: acquiring multimodal medical data of the target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; inputting the multimodal medical data into a multimodal feature extraction model, and outputting multimodal feature data related to the prediction of intracranial atherosclerotic stenosis (ICAS) stroke risk; inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model, and outputting the ICAS stroke risk prediction result of the target object.

[0122] Storage unit 1720 may include readable media in the form of volatile storage units, such as random access memory (RAM) 17201 and / or cache memory 17202, and may further include read-only memory (ROM) 17203.

[0123] Storage unit 1720 may also include a program / utility 17204 having a set (at least one) program module 17205, such program module 17205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0124] Bus 1730 can represent one or more of several types of bus structures, including memory cell bus or memory cell controller, peripheral bus, graphics acceleration port, processing unit, or local bus using any of the multiple bus structures.

[0125] Electronic device 1700 can also communicate with one or more external devices 1740 (e.g., keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with electronic device 1700, and / or any device that enables electronic device 1700 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 1750. Furthermore, electronic device 1700 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 1760. As shown, network adapter 1760 communicates with other modules of electronic device 1700 via bus 1730. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 1700, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0126] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0127] Based on the same inventive concept, this disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the medical data processing method described above. Since the principle by which this computer-readable storage medium solves the problem is similar to that of the above-described method embodiments, the implementation of this computer-readable storage medium embodiment can refer to the implementation of the above-described method embodiments, and repeated details will not be elaborated further.

[0128] More specific examples of computer-readable storage media in this disclosure may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0129] In this disclosure, a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting a program for use by or in connection with an instruction execution system, apparatus, or device.

[0130] Optionally, the program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0131] In practical implementation, program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0132] Based on the same inventive concept, this disclosure also provides a computer program product, comprising: a computer program or instructions, wherein the computer program or instructions, when executed by a processor, implement the medical data processing method of any one of the above method embodiments. Since the principle by which this computer program product embodiment solves the problem is similar to that of the above method embodiments, the implementation of this computer program product embodiment can refer to the implementation of the above method embodiments, and repeated details will not be elaborated further.

[0133] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units. Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0134] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0135] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A medical data processing method, characterized in that, include: Acquire multimodal medical data of the target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; The multimodal medical data is input into a multimodal feature extraction model, which outputs multimodal feature data related to the prediction of stroke risk in intracranial atherosclerotic stenosis (ICAS). The multimodal feature data is input into a pre-trained ICAS stroke risk prediction model, and the ICAS stroke risk prediction result of the target object is output.

2. The medical data processing method according to claim 1, characterized in that, The multimodal medical data includes: multimodal medical data of the target object in different time periods; The process involves inputting the multimodal medical data into a multimodal feature extraction model and outputting multimodal feature data related to the prediction of stroke risk in intracranial atherosclerotic stenosis (ICAS). This includes: inputting multimodal medical data of the target object at different time periods into the multimodal feature extraction model and outputting multimodal feature data of the target object at different time periods; and fusing the same-modal feature data of the target object at different time periods to obtain the multimodal fusion feature of the target object. The process of inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result of the target object includes: inputting the multimodal fusion features of the target object into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result of the target object.

3. The medical data processing method according to claim 1, characterized in that, After inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result for the target object, the method further includes: Based on the ICAS stroke risk prediction results of the target object, a personalized treatment plan is generated for the target object, wherein the personalized treatment plan includes: a treatment plan for the target object in at least one treatment stage; The treatment feedback results of the target object in the first treatment stage are obtained, and the treatment plan of the target object in the second treatment stage is updated according to the treatment feedback results of the target object in the first treatment stage, wherein the first treatment stage is the previous treatment stage of the second treatment stage.

4. The medical data processing method according to claim 1, characterized in that, After inputting the multimodal feature data into a pre-trained ICAS stroke risk prediction model and outputting the ICAS stroke risk prediction result for the target object, the method further includes: The ICAS stroke risk prediction results for the target object are visualized.

5. The medical data processing method according to any one of claims 1 to 4, characterized in that, Acquire multimodal medical data of the target object, including: Based on a pre-trained medical big language model, one or more prompts are generated according to pre-determined ICAS stroke risk prediction indicators. The medical big language model is obtained by fine-tuning the big language model using data from the medical field. The prompts are used to prompt the user to input the indicator data corresponding to each reference indicator. Acquire multimodal medical data input by the user based on various prompts.

6. A medical data processing system, characterized in that, include: The clinical medical data module is used to provide clinical medical data for the target population. The medical imaging data module is used to provide medical imaging data for the target object; The biomarker data module is used to provide biomarker data for the target object; The data processing module is used to input the data provided by the clinical medical data module, the clinical medical data module and the biomarker data module into the multimodal feature extraction model, output multimodal feature data related to ICAS stroke risk prediction, and input the multimodal feature data into the pre-trained ICAS stroke risk prediction model to output the ICAS stroke risk prediction result of the target object.

7. A medical data processing device, characterized in that, include: A multimodal medical data acquisition module is used to acquire multimodal medical data of a target object, wherein the multimodal medical data includes at least one of the following: clinical medical data, medical imaging data, and biomarker data; The multimodal feature extraction module is used to input the multimodal medical data into the multimodal feature extraction model and output multimodal feature data related to the prediction of stroke risk in intracranial atherosclerotic stenosis (ICAS). The prediction module is used to input the multimodal feature data into a pre-trained ICAS stroke risk prediction model and output the ICAS stroke risk prediction result of the target object.

8. An electronic device, characterized in that, include: processor; as well as Memory for storing the executable instructions of the processor; The processor is configured to execute the medical data processing method according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the medical data processing method according to any one of claims 1 to 6.

10. A computer program product comprising: A computer program or instruction, characterized in that, when executed by a processor, the computer program or instruction implements the medical data processing method according to any one of claims 1 to 6.

Images