Assessing the accuracy of magnetic resonance imaging in identifying early rectal cancers suitable for endoscopic intermuscular dissection

Background Selection of rectal cancers suitable for endoscopic intermuscular dissection (EID) is challenging. We aimed to evaluate whether identification of =1mm of preserved muscularis propria on magnetic resonance imaging (MRI), using a systematic reporting tool (mrSRT), can identify rectal cancers suitable for EID. Methods An expert radiologist trained 12 study radiologists in the use of the mrSRT. The radiologists then assessed a retrospective series of MRIs from 269 consecutive patients with suspected deep submucosal invasive rectal cancer. The primary objective was to determine the diagnostic accuracy of =1mm of preserved muscularis propria on MRI for selecting cases that can be resected with clear margins using EID (invasion limited to the circular muscularis propria [=pT2circ]). Diagnostic accuracy was calculated for the expert radiologist, study radiologists, and a consensus diagnosis. Results After applying the inclusion and exclusion criteria, 244 patient scans were included in the analysis. Histological classification confirmed 18 lesions (7.4%) were noninvasive, 109 (44.7%) were pT1, 56 (23.0%) were pT2circ, 21 (8.6%) were pT2long, 39 (16.0%) were pT3, and 1 (0.4%) was pT4. The overall diagnostic accuracy of =1mm of preserved muscularis propria as a criterion for selection was 80.7% (95%CI 75.2-85.5) for the expert radiologist, 77.5% (95%CI 71.7-82.5) for the trained study radiologists, and 81.6% (95%CI 76.1-86.2) for a consensus diagnosis. Conclusion Use of mrSRT to identify =1mm of preserved muscularis propria on MRI allowed radiologists to assist in appropriate case selection for EID.