Agreement between magnetic resonance imaging and ultrasonography in deep pelvic endometriosis

Deep pelvic endometriosis is the most common cause of chronic pelvic pain and infertility. Guidelines proposed standardized approaches for the diagnosis of deep pelvic endometriosis with ultrasonography and magnetic resonance imaging; however, knowing the reasons for discrepancy is crucial. We aimed to analyze the agreement between ultrasonography and magnetic resonance imaging for deep pelvic endometriosis findings and provide potential pitfalls and reasons for discordant findings.

The study group consists of consecutive patients with deep pelvic endometriosis diagnosed on pelvic (n=1) or transvaginal ultrasonography (n=34) who underwent noncontrast pelvic magnetic resonance imaging. The agreement between the ultrasonography and magnetic resonance imaging was assessed using the prevalence and bias-adjusted kappa statistics. Potential pitfalls and reasons for discordant findings were presented.

The study group consisted of 35 patients with deep pelvic endometriosis. The mean age was 39.5±6.2 years. The most common site of involvement was the rectosigmoid colon (n=34, 97.1%), followed by endometrioma/hemorrhagic cyst (n=32, 91.4%). There was a perfect agreement for endometrioma/hemorrhagic cyst (100%), almost perfect agreement for bladder involvement (PABAK=0.886), and moderate agreement for other sites. The number of uterosacral ligament involvement was lower with ultrasonography than with magnetic resonance imaging. However, due to the impact of gas signals on magnetic resonance imaging imaging, the number and boundaries of rectosigmoid deep pelvic endometriosis were better defined with ultrasonography.

The agreement between ultrasonography and magnetic resonance imaging for deep pelvic endometriosis findings varies according to the sites of involvement. Ultrasonography and magnetic resonance imaging are not standalone diagnostic techniques but are complementary to each other. We provided potential diagnostic pitfalls.