A systematic approach to standardizing the visual appearance of endometriotic lesions for artificial intelligence recognition.
Antoine Netter, Fanny Duchateau, Henrique Abrao et al.
✦ AI-curated · Sources linked
In 30 seconds
This systematic review aimed to standardize the visual classification of endometriotic lesions to enhance artificial intelligence recognition. Analyzing 35 studies with histologically confirmed endometriosis, the research developed an ontology categorizing lesions into 4 classes and 11 subclasses. The positive predictive values (PPV) for lesion types varied significantly, with black lesions showing a PPV of 47%-97%.
Key findings
- The visual ontology categorized endometriosis lesions into 4 classes and 11 subclasses.
- Positive predictive values for black lesions ranged from 47% to 97%.
- Red lesions had a positive predictive value between 33% and 100%.
- Nonspecific lesions like adhesions had a lower positive predictive value of 16% to 50%.
Why it matters
Standardizing the visual appearance of endometriotic lesions is crucial for improving diagnostic accuracy and reproducibility during laparoscopy. This can enhance the development of AI tools for better recognition and classification of endometriosis.