# Endometriosis Imaging Diagnosis Beyond Specialized Centers: A Retrospective Cross-Sectional Study

## TL;DR

This study evaluates how well MRI can detect endometriosis in non-specialized hospitals and finds that severe symptoms are linked to diagnosis, even when MRI results are unclear.

## Contribution

The study introduces a real-world evaluation of MRI accuracy for endometriosis in non-specialized centers and identifies symptom severity as a key predictor of diagnosis.

## Key findings

- Pelvic MRI has high specificity but low sensitivity for deep endometriosis in non-specialized centers.
- Severe symptoms strongly correlate with surgical detection of endometriosis, independent of endometrioma size.
- Endometrioma size alone does not reliably predict surgical detection of deep endometriosis.

## Abstract

Objective: Determine the accuracy of deep endometriosis (DE) and ovarian endometrioma (OE) diagnosis on pelvic MRI in non-specialized tertiary healthcare, taking surgical diagnosis as the reference test. Additionally, we aimed to evaluate symptom severity and the presence and size of endometriomas as predictors of DE detection during surgery.

Methods: This retrospective cross-sectional study assessed endometriosis patients presenting to a tertiary healthcare center over one year (2018-2019). Pelvic MRI data, obtained within six months of surgery, were extracted from radiologic reports. A symptom severity score (SSS) was used to estimate symptom severity.

Results: Pelvic MRI exhibited low sensitivity (75.6%) and high specificity (100%) for DE (n=96). Conversely, OE was diagnosed with high sensitivity (100%) and specificity (93.9%; n=122). Three-way contingency analysis showed that, irrespective of the presence and size of endometrioma, there was a correlation between symptom severity and detection of DE during surgery (OR=6.6, p-value≤0.001; n=287). In a multivariable logistic regression analysis model controlling for endometrioma size, a higher SSS was significantly associated with a higher risk of DE detection during surgery (SSS=1: OR=4.65, p≤0.001; SSS=2: OR=10.66, p≤0.001; n=287). However, when controlling for symptom severity, there was no significant association between the presence and size of endometrioma and surgical detection of DE (endometrioma ≥5 cm: OR=1.57, p>0.05; n=287).

Conclusion: While pelvic MRI effectively identifies DE in this population, caution is advised regarding negative DE MRI results in non-specialized settings. Based on our findings, a high suspicion for DE should be maintained in patients with severe symptoms.