Case · Hospital
Tertiary hospital — XAI diagnostic assist for EU AI Act
// PERSONA
Lee, Healthcare IT lead — tertiary hospital
// INDUSTRY
Hospital / Healthcare
// DATASET
case_02 · imaging_diagnosis_log.csv · 8,200 rows
PROBLEM
The Problem
EU AI Act 2026 mandates mathematical justification for high-risk AI such as diagnostic assist, and Korea's regulator is heading the same way. Existing deep-learning models are black boxes, so clinicians have to explain "why is this patient high-risk?" — but the model cannot.
// DEEP LEARNING MODEL
// EU AI ACT 2026
"고위험 AI는 의사결정 근거 의무"
APPROACH
XimTier Approach
PACS and EMR data are integrated inside an isolated environment, and SHAP-based variable contributions are auto-generated for each diagnosis. The clinician UI shows lines like "this case: variable X raised risk by 0.34", and regulator-ready documents are produced automatically.
// SHAP IMPACT
R² 0.91
나이
BMI
흡연 이력
혈압
가족력
✓ EU AI Act + 한국 AI 기본법 통과
OUTCOMES
Outcomes
OUT / 01
EU AI Act + Korea AI Framework both satisfied
OUT / 02
Clinician decision time −34%
OUT / 03
Patient wait time −18 minutes (average)
OUT / 04
Audit/certification docs auto-generated (−80% manual work)
⚠ Example data — fine-tuned on customer data at deployment.