US 12,230,401 B1
Intelligent decision reasoning method for type-based diagnosis and treatment of cardiovascular disease, device, and product
Hongzhen Cui, Beijing (CN); Meihua Piao, Beijing (CN); Yunfeng Peng, Beijing (CN); Shichao Wang, Beijing (CN); Longhao Zhang, Beijing (CN); Haoming Ma, Beijing (CN); and Xiaoyue Zhu, Beijing (CN)
Assigned to University of Science and Technology Beijing, Beijing (CN); and Peking Union Medical College, Beijing (CN)
Filed by University of Science and Technology Beijing, Beijing (CN); and Peking Union Medical College, Beijing (CN)
Filed on Apr. 3, 2024, as Appl. No. 18/626,099.
Int. Cl. G16H 50/20 (2018.01); G16H 10/60 (2018.01); G16H 20/10 (2018.01); G16H 50/70 (2018.01); G16H 70/40 (2018.01)
CPC G16H 50/20 (2018.01) [G16H 10/60 (2018.01); G16H 20/10 (2018.01); G16H 50/70 (2018.01); G16H 70/40 (2018.01)] 17 Claims
OG exemplary drawing
 
1. An intelligent decision reasoning method for type-based diagnosis and treatment of a cardiovascular disease, comprising:
constructing a risk factor mining model of a cardiovascular disease and a drug attribute association mining model of the cardiovascular disease;
extracting common clinical symptoms, disease types, and medication attribute information of a plurality of types of cardiovascular diseases from cardiovascular data based on the risk factor mining model of the cardiovascular disease and the drug attribute association mining model of the cardiovascular disease, wherein the cardiovascular data comprises: data of an electronic medical record of a cardiovascular patient, a cardiovascular medication guideline, clinical reception information, and expert consensus knowledge;
constructing a clinical diagnostic dataset and a clinical symptom diagnosis knowledge system of the cardiovascular disease through data preprocessing based on the common clinical symptoms, the disease types, and the medication attributes of the plurality of types of cardiovascular diseases;
constructing a type-based auxiliary diagnosis model of the cardiovascular disease based on the clinical diagnostic dataset and the clinical symptom diagnosis knowledge system of the cardiovascular disease, and using the type-based auxiliary diagnosis model of the cardiovascular disease for intelligent decision reasoning;
constructing an association rule based on a reasoning result obtained through the intelligent decision reasoning;
constructing a medical convalescence knowledge base of the cardiovascular disease based on convalescence data of the cardiovascular patient and the association rule;
generating a recommended patient wellness program based on a collaborative filtering algorithm and the medical convalescence knowledge base of the cardiovascular disease; and
treating the cardiovascular disease according to drug items of the recommended patient wellness program;
wherein the constructing a risk factor mining model of a cardiovascular disease and a drug attribute association mining model of the cardiovascular disease comprises:
extracting consensus text data by using a document parsing technology, wherein the consensus text data comprises text data of the electronic medical record of the cardiovascular patient and the cardiovascular medication guideline;
adding a data annotation to the consensus text data to obtain annotated data, wherein the data annotation comprises an entity type, a relationship type, and an attribute type, the attribute type comprises a symptom, a cardiovascular disease type, a risk factor, a medical history, a biochemical indicator, a drug name, a medication type, a mode of administration, a medication frequency, a medication cycle, and a medication dosage;
training a network model based on the annotated data, wherein the network model comprises an input layer, a text information representation and embedding layer, a semantic information modeling layer, a label sequence correction and recognition layer, and an output layer, wherein the input layer is configured to receive a training corpus for the risk factor mining model of the cardiovascular disease and the drug attribute association mining model of the cardiovascular disease, the text information representation and embedding layer comprises a Generalized Autoregressive Pretraining for Language Understanding (XLNet) network model integrating text semantic information and positional information to achieve corpus vectorization, the semantic information modeling layer comprises a Bi-directional Long Short-Term Memory (BiLSTM) network model comprising a forward Long Short-Term Memory (LSTM) layer capturing contextual information and a reverse LSTM layer capturing reverse information, the label sequence correction and recognition layer comprises a Conditional Random Field (CRF) network model configured to perform corrective recognition on a predicted label sequence output by the semantic information modeling layer, and obtain an optimal solution based on a probability relationship between adjacent labels, and the output layer is configured to generate a label sequence prediction result corresponding to a text corpus of the input layer; and
separately using a trained network model as the risk factor mining model of the cardiovascular disease and the drug attribute association mining model of the cardiovascular disease.