Fusion 3D T1/T2 MRI for diagnosing pelvic deep infiltrating endometriosis: a non-inferiority study

Magnetic resonance imaging (MRI) is an essential non-invasive technique for diagnosing deep infiltrating endometriosis (DIE). Published studies have explored the value of high-resolution fat-suppressed 3D T1-weighted (T1w) and 3D T2-weighted (T2w) MRI in this topic. However, research on the fusion of these sequences in a merged 'PET-scanner-like' multiplanar reconstruction (Fusion 3D T1/T2) is still lacking. Our primary objective was to assess whether the diagnostic accuracy of Fusion 3D T1/T2 was non-inferior to the state-of-the-art MRI protocol for diagnosing DIE. This standard protocol consists of 2D T2w images in various planes and axial T1w images displayed on a four-window screen (STANDARD). The non-inferiority margin was set at 2.5 %. Fifty-nine pelvic MRI examinations for suspected endometriosis were interpreted independently by two radiologists with different experience levels (R#1 and R#2), randomly with both protocols. Results were compared with a consensus opinion from an expert panel. Interpretation time and inter-reader agreement were also evaluated. Using the Fusion 3D T1/T2 protocol, R#1's diagnostic accuracy for DIE was 94.74% (95%CI: 92.48-96.47) and R#2's was 98.68% (95%CI: 97.31-99.47), both of which were non-inferior to the STANDARD protocol (R#1, 92.48%, 95%CI: 89.90-94.57; R#2, 96.43%, 95%CI: 94.48-97.84). The Fusion 3D T1/T2 protocol reduced interpretation time by 24.5 % (P < 0.001) compared to the STANDARD. Additionally, it improved inter-reader reproducibility, with a moderate level of agreement (kappa = 0.72; 95 % CI: 0.63-0.81), compared to the slight agreement (kappa = 0.47; 95 % CI: 0.34-0.59) observed with the STANDARD protocol. In conclusion, Fusion 3D T1/T2 demonstrated non-inferior diagnostic accuracy compared to the current MRI standard protocol. It also improved reproducibility and reduced interpretation time, suggesting its potential as a valuable tool for diagnosing pelvic DIE.