ML/DL models show high diagnostic accuracy for endometriosis but lower specificity limits clinical useMachine learning models show promise in diagnosing endometriosis

Frontiers in Medicine Published October 16, 2026 Study authors: Jiayi Guo, Xue Liu, Huiying Zhang, Shaobin Wei DOI ↗ Editorial oversight: Dr. Sofia Müller, MD · Lifespan & Whole-Person Care

AI-generated summary of the cited source, checked by automated accuracy review. How we work

Key Takeaway

Consider ML/DL models as promising but not yet validated for routine endometriosis diagnosis.

This meta-analysis evaluated the diagnostic accuracy of machine learning (ML) and deep learning (DL) models for endometriosis. The authors pooled sensitivity, specificity, and summary receiver operating characteristic (SROC) curve data. No details on the included studies' populations, sample sizes, or settings were reported in the source.

The pooled sensitivity was 0.90 (95% CI 0.86 to 0.93), specificity 0.75 (95% CI 0.64 to 0.84), and SROC-AUC 0.92 (95% CI 0.89 to 0.94). In a modeled scenario, the probability of disease after a positive prediction increased from 10% to 29%, and after a negative prediction decreased to 1%. The authors reported that results were robust.

Limitations noted by the authors include that multicenter design and DL were associated with lower sensitivity. They also stated that large multicenter prospective studies and multimodal models are needed to enable clinical translation. No adverse events or safety data were reported, as expected for diagnostic accuracy studies.

In practice, these findings suggest that ML/DL models may serve as promising non-invasive auxiliary diagnostic tools, but the lower specificity and the need for prospective validation mean they are not yet ready to replace standard diagnostic pathways.

How this fits prior evidence

This meta-analysis extends prior coverage of endometriosis pathophysiology, which focused on inflammation-fibrosis circuits, immune-fibrotic frameworks, and Hippo/YAP1 signaling as non-hormonal targets, by shifting attention to diagnostic technology. It also contrasts with a prior protocol for ethanol sclerotherapy, which addressed a therapeutic procedure, and with evidence of low patient awareness of non-surgical diagnosis and preoperative imaging for bowel endometriosis. The high pooled sensitivity (0.90) and SROC-AUC (0.92) suggest ML/DL could complement imaging and clinical assessment, but the specificity of 0.75 and the need for prospective validation temper immediate clinical adoption.

Living with endometriosis can be a long and frustrating journey because the condition is often hard to spot. New research looks at how machine learning and deep learning models, which are types of advanced computer programs, can help doctors identify the disease more reliably.

The analysis found that these computer models have high diagnostic accuracy. Specifically, they showed a sensitivity of 0.90, which means they are good at correctly identifying people who have the condition. They also showed a specificity of 0.75. When the computer predicted a positive result, the chance of actually having the disease rose from 10% to 29%. When it predicted a negative result, the chance of having the disease dropped to 1%.

While these results are robust, there is still work to do. Some designs showed lower sensitivity, and more large, multi-center studies are needed before these tools can be used in every clinic. For now, these models show promise as a non-invasive way to help doctors make better calls.

What this means for you:

Machine learning models show high accuracy in identifying endometriosis and could help doctors provide faster diagnoses.

Common questions

How accurate are these computer models at finding endometriosis?

The models showed high diagnostic performance. They had a sensitivity of 0.90 and a specificity of 0.75. When the model gave a positive prediction, the probability of having the disease increased from 10% to 29%. When it gave a negative prediction, the probability dropped to 1%.

What are the limitations of using machine learning for diagnosis?

While the results were robust, some designs showed lower sensitivity. Because these are still developing tools, larger studies involving many different centers are needed before they can be fully used in standard clinical practice.

How can these models help patients with endometriosis?

These models are promising because they offer a non-invasive way to help doctors diagnose the condition. They can act as an extra tool to help doctors make more accurate decisions for their patients.

Study Details

Study typeMeta analysis

EvidenceLevel 1

PublishedOct 2026

View Original Abstract ↓

To assess the diagnostic accuracy of machine learning (ML) and deep learning (DL) models for endometriosis. A systematic review and meta-analysis were conducted in accordance with PRISMA guidelines (PROSPERO: CRD420251231469). Databases were searched through November 24, 2025. Twenty-two eligible studies were included. Risk of bias was assessed using QUADAS-2. Pooled analysis was conducted in Stata 18.0. Pooled sensitivity was 0.90 (95% CI 0.86–0.93), specificity 0.75 (95% CI 0.64–0.84), and SROC-AUC 0.92 (95% CI 0.89–0.94). A positive prediction increased disease probability from 10% to 29%, whereas a negative prediction reduced it to 1%. Multicenter design and DL were associated with lower sensitivity. Results were robust. ML/DL models demonstrate high diagnostic performance and are promising for non-invasive auxiliary diagnosis. Large multicenter prospective studies and multimodal models are needed to enable clinical translation. Systematic Review Registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD420251231469, identifier CRD420251231469

ML/DL models show high diagnostic accuracy for endometriosis but lower specificity limits clinical useMachine learning models show promise in diagnosing endometriosis

How this fits prior evidence

Common questions

Study Details

Platinum agents, PARP inhibitors, and immune checkpoint inhibitors increase pCR rates in early-stage TNBC

New data shows specific drug combinations improve triple-negative breast cancer outcomes

Clinical research that matters. Delivered to your inbox.

ML/DL models show high diagnostic accuracy for endometriosis but lower specificity limits clinical useMachine learning models show promise in diagnosing endometriosis

How this fits prior evidence

Common questions

More on Endometriosis

Study Details

Platinum agents, PARP inhibitors, and immune checkpoint inhibitors increase pCR rates in early-stage TNBC

New data shows specific drug combinations improve triple-negative breast cancer outcomes

Clinical research that matters. Delivered to your inbox.

Related in OB/GYN & Women's Health

From Other Specialties