Title: DUCG-aided general clinical diagnosis enabling primary clinicians to diagnose diseases like experts and know why: Real applications

Abstract

DUCG(Dynamic Uncertain Causality Graph) is a new AI model to graphically represent domain uncertain causal knowledge and make probabilistic reasoning with natural interpretabilities. This presentation will show online how DUCG works to guide primary clinicians to make clinical diagnoses under 43 chief complaints covering more than 1000 diseases, including how to collect clinical information and make what medical checks, step by step according to the condition of a primary hospital or clinics. The 43 chief complaints include: dizzy, headache, nasal congestion, epistaxis, sore throat, jaundice, dysphagia, cyanosis, cough and expectoration, dyspnea, neck waist and back pain, hemoptysis, lymphadenopathy, chest pain, palpitation, hematemesis, arthralgia, abdominal pain, nausea and vomiting, numbness of limbs, edema, bloody stool, constipation, rash, fever, anemia, obesity, emaciation, child fever, diarrhea, syncope, abdominal distention, related disease of department of gynecology (four chief complaints), lower urinary tract symptoms(including frequent urination, urgency, pain, dysuria, polyuria, gross hematuria and urinary leakage).In total, the differential diagnostic precision verified by third-parties for every chief complaint is more than 95%, in which the precision for every disease is no less than 80%. More than 200,000 real application cases were performed in Jiaozhou city and Zhongxian county, China. In which, only 8 diagnoses were incorrect due to the imperfectness in DUCG knowledge bases. After correcting DUCG knowledge bases, incorrect diagnoses were no long found. Statistics in the real world shows that DUCG can increase the ability of primary clinicians to diagnose diseases several times overthat without DUCG.