The Study Of Intelligent Assistant Medical Diagnosis And Medical Question-answer System Technologies

Smart healthcare has always been a research hotspot in the development of artificial intelligence.Constrained by regional economic development,medical resources are unevenly distributed across regions.Famous doctors often work in economically developed areas,and many basic community hospitals lack excellent doctors,resulting in most patients not being able to receive an efficient and accurate diagnosis and treatment.Therefore,the importance of the development of intelligent assisted medical technology for doctors is very urgent.At present,Bayesian probabilistic inference technology and neural network technology have achieved good results.Bayesian probabilistic reasoning has good interpretability and accuracy,but the computational complexity is too high to make it possible to perform large-scale reasoning.Neural network technology can have high accuracy through self-learning,but its inexplicability is prohibitive.This article uses a combination of probabilistic reasoning technology and neural network technology to improve the interpretability and accuracy of the model,and considers to provide doctors with diagnostic recommendations and "diagnosis of disease from symptoms" and "consultation from questions." Questions and answers.Specifically,the main research contents of this article are as follows:(1)An active interactive intelligent assistant diagnosis model is proposed.The model infers the probability of each disease according to the patient's symptoms and recommends the optimal test item,and iterates repeatedly until the diagnosis result is given.Specifically,a Partial-Bayesian Network(PB-Net)algorithm is designed in the diagnostic module to infer the probabilities of various diseases.The recommendation module contains a novel Medical Tests Recommendation Model(MTR)based on the reinforcement learning to choose an optimal medical test in each interactive state.Finally,in experiments,we proved that our model can improve the accuracy and efficiency of disease diagnosis.(2)A Dependent Multilevel Interaction Model(DMI)based on hierarchical interaction is proposed to predict the correlation between "question" sentences and "answer" sentences.Specifically,the model first designs a Single Interaction Unit(SIU)including a combining attention mechanism and a comparison module,and then a multilevel interaction structure is constructed to deepen the degree of interaction between the question and answer sentences to extract features from multiple aspects.Besides,parameters will be transferred between different levels to enhance dependencies and reduce information redundancy.In the experiments,the model achieves the highest scores on the public datasets SciTail and SNLI,which demonstrate the advancement of our model.(3)The above models and algorithms are applied to real scenarios,and an intelligent assistant diagnosis and medical question answering system for doctors is constructed.The system can assist doctors in giving probabilities of disease inference and recommendations for medical tests during each step of a diagnosis.Simultaneously,doctors can ask questions in the QA module to search for relevant answers.