Preoperative MRI-based Deep Learning Reconstruction and Classification Model for Assessing Rectal Cancer

Motivation: Excellent magnetic resonance imaging (MRI) image quality is a prerequisite for accurate diagnosis of rectal cancer (RCa). Goal(s): To explore the feasibility of image enhancement by deep learning reconstruction (DLR) and its improvement in TN stage of patients with RCa. Approach: Images of T2WI, DWI, and contrast-enhanced T1WI from patients with pathologically diagnosed RCa were retrospectively processed with and without DLR and assessed by five readers. Qualitative and quantitative image quality, TN stage, and diagnositc performance were assessed. Results: DLR and classification models could enhance the MR image quality and improve the diagnositic performance of TN stage for patients with RCa. Impact: Deep learning reconstruction could improve the image quality of rectal MRI and enhance the diagnostic performance for the TN stage of rectal cancer, which could be used to promote visualization and diagnostic performance in patients with rectal cancer.