Radiomics and artificial intelligence for imaging-based non-invasive diagnosis and phenotyping of endometriosis: A systematic review

Deku Liang, Li He
In: Journal of Radiation Research and Applied Sciences · 2026 · vol. 19(3) , pp. 102601 · doi:10.1016/j.jrras.2026.102601 · W7196945062
Review OA: gold CC0
🔓 Open OA copy View on OpenAlex View at publisher
⚙ AI-generated summary by gemini-2.5-flash-lite, 2026-08-11 ⓘ

Radiomics and AI applied to MRI/ultrasound show potential for non-invasive endometriosis diagnosis and phenotyping, with high diagnostic performance in some tasks but requiring further validation.

One-sentence paraphrase of the abstract; not a substitute for reading it. No clinical advice. How this works

Abstract

Background Endometriosis affects approximately 15% of women of reproductive age and is associated with diagnostic delays of up to 12 years due to reliance on surgical confirmation. Radiomics and artificial intelligence (AI) applied to medical imaging offer a pathway toward non-invasive, biologically informed diagnosis, but the evidence base has not been systematically appraised. Methods A systematic search for PubMed/MEDLINE, Scopus and Web of Science databases was performed following the PRISMA 2020 criteria. All original studies involving the use of MRI and/or ultrasound for endometriosis diagnosis, classification, segmentation or phenotyping, using radiomics, were included. Data were extracted according to a pre-specified form and methodological quality was evaluated with QUADAS-2. Results There were 18 studies (2020-2026) that included 4381 patients, mostly with a single-center, retrospective design. In ovarian lesion differentiation, deep learning achieved high Area under the curve (AUC) values (up to 0.987); for non-invasive rASRM staging, the accuracy was 87.8% and for detection of pouch of Douglas (POD), AUC values reached up to 96.5%. Across the reader study experiments reported, AI assistance was generally associated with improved sensitivity and inter-observer agreement among radiologists. The risk of bias was variable, with patient selection and reference standard being the most common areas of concern. There was a small number of studies that were externally validated. Given substantial clinical and methodological heterogeneity across studies, findings were synthesized narratively rather than pooled meta-analytically, and reported performance figures should be interpreted as individual study estimates rather than generalizable pooled measures. Conclusions Radiomics and AI show potential for diagnosis for various phenotypes of endometriosis, but their clinical application will need prospective multicenter validation, standardization of protocols, and extending this to less represented phenotypes such as superficial peritoneal disease and monitoring treatment response.

My notes (saved in your browser only)

Citation neighborhood (2-hop)

Papers in the corpus that this work cites (lower rings, blue) and that cite this one (upper rings, green). Dot size scales with the paper's in-corpus citation count — bigger dot = more influential within the endo/adeno field. Click a dot to open that paper. Outer rings show 2-hop neighbours — papers reached through the immediate citers/citees. [ collapse to 1-hop ]

References (35)

Source provenance

openalex
last seen: 2026-10-15T06:00:39.803327+00:00
License: CC0 · commercial use OK