Feasibility of accelerated 3D T2-weighted bladder MRI using the deep learning-constrained compressed sensitivity encoding technique (SENSE)

Motivation: To compare the application value of AI algorithm in 3DT2WI of bladder cancer Goal(s): This study aimed to evaluate the clinical value of deep-learning algorithm reconstruction in 3DT2WI for bladder cancer. Approach: Sixty-seven patients with bladder cancer (MIBC /NMIBC= 12/42) underwent high-resolution 3DT2WI with and without the deep-learning reconstruction algorithm (DLA) at 3.0T. The pathological results were used as the gold standard for diagnostic evaluation. Results: Regarding the diagnostic efficiency, the AUCs of the two 3DT2WI-DLA, compressed SENSE(CS), and the compressed 32.5% scan time, were 0.895 and 0.888, respectively. In addition, 3DT2WIDLA had higher scores for image quality than 3DT2WI. Impact: DLA-CS cloud helped high-resolution 3DT2WI decrease the scantime, meet the diagnostic requirements, improve image quality and decrease artifacts in bladder cancer.