Architecture of an Expert System to Support Diagnostic Decisions for Hereditary Diseases

The number of genetically related diseases is over 7000, of which most are extremely rare. Early diagnosis of these diseases is associated with several difficulties, including a variety of clinical forms, polymorphism (multivariation) of phenotypic manifestations, and lack of personal experience in observing patients with these pathologies. The use of physician-assisted computer systems may allow to overcome these difficulties. For this purpose, an expert system was developed to support diagnostic decisions in hereditary lysosomal diseases. Knowledge extraction took place in two stages—from literature sources and from experts. The knowledge base is implemented on a cloud platform in the form of an ontological network. The mathematical model of the disease allows a complex assessment of the signs based on, the modality coefficient and confidence measures of manifestation and degree of expression suggested by the experts. The comparative analysis algorithm compares the new case to the reference variants of the known clinical forms of the integral model and then ranks the hypotheses put forward. The explanation block allows to present the data that served as a basis for the hypothesis because of the features: confirming the hypothesis, missing to confirm the hypothesis, or irrelevant to the diagnosis. The results of clinical testing of the system showed high (above 88%) efficiency of differential diagnosis.