Design And Implementation Of Computer Aided Diagnosis System Based On Satisfiability Problem Solving

In recent years,the concept of “intelligent healthcare”has gradually emerged.This paper focuses on the computer-aided diagnosis scenario in intelligent healthcare,that is when a computer can automatically diagnose disease based on patient’s symptom information.For the time being,the issue of computer-aided diagnosis is solved with big data + deep learning method being involved,that is,deep learning solution is used to learn the association model of diseases and symptoms from a big data of electronic medical records,with further diagnosis based on the patient’s symptoms.The distinctive feature of this method is that a large amount of data can contain almost all common diseases and symptoms.But the out coming issues are also very clear.First,this method relies heavily on medical records,which may not cover rare diseases.Second,the results received from the deep learning lack correct interpretation.And finally,there is a problem of data reliability as there may occur errors in the medical records.Unfortunately,big data cannot provide a solution to detect the errors from the correct data.Focusing on the problems of big data solution,this paper proposes a new computer-aided diagnosis method: first,extract medical knowledge to create a valid knowledge base(KB)from trusted medical resources(such as medical books,professional medical knowledge bases),and verify such a knowledge base to ensure its reliability.Second,encode medical knowledge into propositional logical formulas,and combine these logical formulas to build instances of satisfiability problem.Finally,the SMT solver is used to find plausible answers for these instances,and on the basis of the found solutions,diagnose a patient.Because we mainly form knowledge base from credible knowledge resources(such as medical books)and have a review process,the reliability and coverage of the received data are better than the one received from medical records.And due to the fact that the medical KB that we possess is based on causal relationships retrieved from the medical books and the fact that we use logical reasoning to simulate diagnosis procedure,our diagnosis results are supported by knowledge from the medical books,ensuring the possibility to interpret the diagnosis results correctly.In addition,we also propose a knowledge error correction scheme which can calibrate the results received in the process of knowledge extraction.The main contributions of this paper are as follows:(1)Creation of a medical knowledge baseIn order to implement the computer-aided diagnosis system described in this paper,we first build a medical knowledge base that meets our requirements.According to the pre-designed medical knowledge model,the necessary data is extracted from a trusted knowledge source,then expressed as a medical knowledge graph and stored in the graph database Neo4 j for subsequent algorithms usage.(2)Computer-aided diagnosis coreBased on the well-built knowledge base,this paper implements the computer-aided diagnosis core.It can extract relevant medical knowledge from the knowledge base on the basis of the patient’s symptoms and encode it into logical formulas.And it then combines these logical formulas to construct instances of the satisfiability problem.Finally,it uses the SMT solver to calibrate its satisfiability in order to perform diagnosis.(3)Knowledge error correction schemeAlthough we extract and review knowledge from trusted sources,due to the disadvantages of manual or automatic extraction algorithms,there still may be errors in the final acquired knowledge.Because this paper uses medical knowledge-based diagnose methods rather than big data solution,erroneous knowledge will greatly affect the accuracy of the final diagnosis.Therefore,such an instance must be corrected.This paper is based on the unsatisfiable core of satisfiability problem to propose a way to correct the contradictory knowledge in the knowledge base.(4)Computer-aided diagnosis system for practical applicationAll the schemes in this paper are not only theoretical,but are also implemented in computer-aided diagnosis systems and have been tested in cooperative hospitals.