대한신장학회

논문분류	춘계학술대회 초록집
제목	Deep Learning-Based Anomaly Detection Using Creatinine and eGFR Time-Series Data to Identify Kidney Injury in ICU Patients
저자	Min Woo Kang
출판정보	2025; 2025(1):
키워드	Acute Kidney Injury, Deep Learning, Anomaly Detection, Creatinine Time-Series
초록	Current acute kidney injury (AKI) diagnosis relies primarily on discrete creatinine changes. However, these static criteria may miss subtle, dynamic changes in kidney function, particularly in rapidly changing intensive care unit (ICU) environments. Deep learning-based anomaly detection may address these limitations by analyzing patient-specific temporal patterns of creatinine. This study analyzed intensive care unit (ICU) patients from the Medical Information Mart for Intensive Care (MIMIC)-IV and eICU databases, excluding those with end-stage kidney disease or prior kidney replacement therapy (KRT). Serum creatinine and estimated glomerular filtration rate collected at 24-hour intervals—from 24 hours before ICU admission until discharge, death, or KRT initiation—were linearly interpolated, with physiologically permissible noise added. Kidney injury by anomaly detection using an Anomaly Transformer model was defined when creatinine increased at the final measurement with an anomaly score above the 5th percentile threshold. Performance for predicting KRT and mortality was internally and externally validated. A total of 61,373 ICU patients (81,876 admissions) from MIMIC-IV generated 381,700 seven-day datasets (80% training, 5% validation, 15% test). External validation utilized 494,684 datasets from 124,348 patients (140,237 admissions) in the eICU database. Kidney injury by anomaly detection occurred in 3.91% of test cases, slightly higher than AKI stage ≥2 (3.79%). Anomaly detection outperformed AKI stage ≥2 criteria in predicting KRT initiation at 96h (test F1-score: 0.265 vs. 0.206; external F1-score: 0.161 vs. 0.132). Combining both criteria improved mortality prediction at 96h (test F1-score: 0.175 vs. 0.153 for AKI stage ≥2 alone). Anomaly scores were significantly higher in patients with AKI stage ≥2 and adverse outcomes, consistently across datasets. Deep learning-based anomaly detection effectively identifies kidney dysfunction, offering an improved criterion for predicting key clinical outcomes in ICU patients. This novel approach may serve as a complementary diagnostic tool in AKI research and clinical ICU practice, pending further validation.
원문(PDF)	PDF 원문보기

간행물 검색

대한신장학회 회원관리 규정