Analysis of Deep Learning-based Phase Correction Applied to Single-Shot rFOV Diffusion Images of the Prostate at 1.5T

Motivation: Deep learning reconstruction algorithms of diffusion datasets of the prostate have enabled improved IQ/increased SNR/reduced scan times, but their application for optimal data combination remains an active area of investigation. Goal(s): Apply a new DL-based phase correction (DLPC) method to improve image quality of reduced field of view diffusion images of the prostate at 1.5T. Approach: 21 patient datasets were acquired and reconstructed with a commercially available DL product and a new DLPC model. Post-processed high b-value and ADC images were assessed qualitatively and quantitatively. Results: DLPC reconstruction resulted in further SNR improvements/reduced image noise/improved image sharpness while maintaining clinically acceptable ADC values. Impact: Improved 1.5T diffusion image quality using DL with phase correction may improve prostate cancer diagnosis and staging. Further work is needed to determine if this technology provides similar improvements image at 3T and may be translated to other sequences.