Accelerated 3D high-resolution T2-weighted breast MRI with deep learning constrained compressed sensing, comparison with conventional T2-weighted sequence on 3.0 T

Abstract

PurposeTo evaluate the feasibility of isotropic 3D high-resolution T2-weighted imaging (T2WI) MRI sequences and compare the images reconstructed by integrating artificial intelligence-compressed sensing (AI-CS), compressed sensing (CS), and conventional 2D T2WI sequences for quality. Materials and methodsFifty-two female patients (ages: 26–80 years) with suspected breast cancer were enrolled. They underwent breast MRI examinations using three sequences: conventional T2WI, CS 3D T2WI, and AI-CS 3D T2WI. Image quality, signal-to-noise ratio (SNR), contrast-to-noise ratio, tumor volume, and maximal tumor diameter were compared using the Friedman test. Image quality was scored on a 5-point scale, with 1 indicating nonassessable quality and 5 indicating excellent quality. Tumor volume and maximal tumor diameter were compared based on AI-CS 3D T2WI (slightly high signal), conventional T2WI, and dynamic contrast-enhanced (DCE) sequences. ResultsAll three T2WI were successfully performed in all patients. 3D CS and AI-CS were significantly better than conventional T2WI in terms of lesion conspicuity and morphology, structural details, overall image quality, diagnostic information for breast lesions, and breast tissue delineation (P < 0.001). The SNR of conventional T2WI was significantly higher for 3D T2WI sequences. The contrast-to-noise ratio was significantly higher for AI-CS 3D T2WI than for conventional T2WI sequence. There was no significant difference in tumor volume between DCE (8.08 ± 16.51) and AI-CS 3D T2WI (8.25 ± 16.29) sequences and no significant differences in tumor diameter among DCE, AI-CS 3D T2WI, and conventional T2WI sequences. ConclusionIsotropic-resolution 3D T2WI sequences can be acquired using AI-CS while maintaining image quality and diagnostic value, which may pave the way for isotropic 3D high-resolution T2WI for clinical application.