Body / AbdominalAI / InformaticsResearch
EndoFusion: Unpaired multi-modal AI detects endometriosis signs on TVUS and MRI
Artificial intelligence in medicine4d ago
EndoFusion, an unpaired multi-modal AI framework, detected endometriosis signs (POD obliteration, bowel nodules) on TVUS and MRI with average AUC 0.827 (95% CI 0.790–0.861).
- EndoFusion enables detection of pouch of Douglas obliteration and bowel nodules from either transvaginal ultrasound or MRI alone, without needing paired exams.
- The framework uses dynamic mutual knowledge distillation and label correlation modeling to transfer knowledge across modalities and signs.
- Single-institution proof-of-concept study; external validation not yet performed.
Automated summary
RadPigeon summaries are original and for information only. They are not clinical advice.