Design: Retrospective observational and analytical study carried out in a specialist endometriosis center in Belo Horizonte, Brazil.

Setting: All patients were submitted to pelvis MRI after institutional protocol for bowel preparation and the results were compared with surgical findings.

Patients or Participants: A total of 69 patients, aged between 18 and 50 years old, underwent pelvis MRI and laparoscopic (LPS) treatment for suspected endometriosis between March 2021 and October 2023.

Interventions: As a retrospective observational study, no interventions were necessary.

Measurements and Main Results: Retrospective revision of pelvis MRI assessed for endometriosis at all sites used in the 2021 AAGL Endometriosis Classification and classified patients into AAGL-MRI stages. Results were compared with reference-standard LPS (AAGL-LPS) staging. The AAGL-MRI and AAGL-LPS stage were concordant in 73,91% of cases. The highest agreement was observed in AAGL-LPS stage 4 (79,7%). Endometriosis was most accurately diagnosed in left ovary (WK 0,481), vaginal cul-de-sac (WK 0,501), rectum/sigmoid colon (WK 0,470), retrocervical (WK 0,447), and bladder (WK 0,534). The worst agreement was seen in lesions at superficial peritoneum (WK 0,008), tubes (WK 0,085 and WK 0,145, right and left respectively), ureters (WK 0,016 and WK 0,026, right and left respectively), rectovaginal septum (WK 0,184), and small bowel/cecum (WK 0,156).

Conclusion: The estimation of the 2021 AAGL Endometriosis Staging by pelvis MRI appears to be feasible, demonstrating moderate to high accuracy in most anatomical sites. The assessment of the bladder, cul-de-sac, ovaries, rectum/sigmoid colon and retrocervical region seems to be very well carried out through the AAGL-MRI classification. However, in superficial peritoneum, tubes and ureters, the classification could be limited. More prospective, multicentric studies are necessary for an adequate validation of a classification of endometriosis applied to MRI.