한의 학파별 처방 추천을 위한 기계 학습 데이터 구축 프로세스 설계 및 시스템 개발

Abstract

It takes a lot of time, manpower, and budget to build high-quality learning data on a large scale, which is a key element for implementing artificial intelligence (AI) in various fields. Unlike Western medicine, in the case of Korean medicine, the characteristics are different according to the school. The machine learning data built by experts from representative schools can be reservoir of clinical knowledge and be used for AI recommendation systems. Therefore, in this study, a representative Korean medicine school was selected, a machine learning data construction process was designed, and a related system was developed. First, three Korean medicine schools were selected in consideration of its utilization in the clinical field and representativeness of Korean medicine treatment. We designed a process composed of collecting patient information through clinical research and build machine learning data through case reviews by clinical experts in each school. To support this process, an electronic case report form (eCRF) system and a symptom/prescription input system were designed and developed. Based on the simulation of patient case evaluation by experts from the three previously selected schools and advisory opinions, case items consisting of 9 groups and 59 detailed items were selected. According to the case items, the eCRF system was designed and developed. In addition, a symptom/prescription input system was designed and developed to enter the main symptoms, diseases, and prescriptions derived during the case review process. The two systems designed in this study can function as a new platform to analyze major symptoms/signs and prescriptions identified by different Korean medicine schools, and can be used to build machine learning data for AI recommended by schools.