Body / AbdominalEmergencyAI / InformaticsResearch
Dual-phase CT radiomics predicts AKI after cardiac arrest with AUC 0.78, but stability concerns persist
Radiology AI literature (PubMed)2w ago
In a retrospective 155-patient study, dual-phase CT radiomics predicted acute kidney injury after cardiac arrest with an AUC of 0.783 in the test set, but feature selection was unstable and repeated resampling median AUC was 0.663 — early and not yet externally validated.
- Retrospective single-center study of 155 non-traumatic OHCA patients (30.3% developed AKI) using admission dual-phase CT radiomics.
- Test-set AUC: radiomics-LR 0.783, kNN 0.778, SVM 0.757; clinical-only model AUC 0.549; combined model didn't improve on radiomics alone (AUC 0.779).
- Repeated resampling stability analysis showed median test-set AUC of only 0.663 (IQR 0.623–0.699), highlighting limited feature-selection stability and the need for external validation before any clinical use.
RadPigeon summaries are original and for information only. They are not clinical advice.