Objectives: This study explores the determinants of medical AI adoption by analyzing 12 global cases (2010–2026) and comparing IBM Watson with Doctronic AI using Rogers' DOI theory. Methods: Cases were selected via a PRISMA review on a two-axis matrix (clinical risk × doctor intervention). The IMDRF SaMD N12 classified risk and FDA intended-use language set interventions; DOI attributes were rated H/M/L from peer-reviewed and regulatory evidence. Results: Performance alone does not explain adoption. The matrix was U-shaped: Gen 1 (high-risk/autonomous) faced barriers, Gen 2 (low-risk/mandatory) entered the market, Gen 3 (high-risk/autonomous) re-entered via institutional support. Trialability (sandboxes, First-N rule) and Observability were decisive, with domain selection as an adoption strategy. Conclusions: Adoption depends on Trialability and Compatibility over performance. Extending Topol's Deep Medicine, it proposes AI Autonomous Delegation consisting of low-risk entry before high-risk scaling.
목차
Ⅰ. 서론 1. 연구의 필요성 2. 연구의 목적 3. 연구의 방법 Ⅱ. 연구방법 1. 의료 AI의 분류 2. 의료 AI의 발전 과정 Ⅲ. 연구결과 1. 의료 AI의 대표 사례 비교 분석 2. 사례 1:왓슨 포 온콜로지 3. 사례 2: Doctronic AI 4. 사례 비교 연구 결과 Ⅳ. 고찰 Ⅴ. 결론 REFERENCES