Nuclear / MolecularBody / AbdominalAI / InformaticsResearch
CAD system with follow-up reference standard accurately classifies 18F-PSMA PET/CT findings in prostate cancer BCR
Radiology AI literature (PubMed)1w ago
A stacking ensemble CAD system achieved 92.5% accuracy, 92.3% sensitivity, and AUROC 0.97 for classifying 18F-PSMA PET/CT findings in prostate cancer biochemical recurrence, using post-therapy follow-up as reference standard. Retrospective, internally validated.
- Retrospective, single-center study of 69 patients (334 findings from 45 with follow-up reference standard), evaluated with stratified 10-fold cross-validation.
- The stacking ensemble model (MLP + random forest) achieved similar accuracy (91.7%) and AUROC (0.97) when expert visual interpretation was the reference standard.
- No external validation; small sample and internal cross-validation may overestimate performance.
RadPigeon summaries are original and for information only. They are not clinical advice.