Deep Learning Enables 60% Accelerated Volumetric Brain MRI While Preserving Quantitative Performance: A Prospective,...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Quantitative Imaging Analysis

Background:

• Volumetric magnetic resonance imaging (MRI) is crucial for quantitative analysis.
• Accelerated MRI sequences aim to reduce scan times but can impact image quality and consistency.
• Deep learning (DL) tools offer potential for enhancing accelerated imaging.

Purpose of the Study:

• To evaluate a vendor-agnostic deep learning tool (SubtleMR) for 60% accelerated volumetric MRI.
• To compare image quality and quantitative analysis consistency against standard-of-care MRI.
• To assess the impact of DL on radiologist perception and clinical classification.

Main Methods:

• Prospective, multicenter study with 40 subjects and 6 MRI scanners.
• Acquisition of standard-of-care and 60% accelerated datasets, with DL processing applied to accelerated scans.
• Quantitative analysis (NeuroQuant), clinical classification, and blinded radiologist review for image quality assessment.

Main Results:

• Accelerated MRI with DL processing was statistically superior in perceived image quality compared to standard-of-care, despite a 60% scan time reduction.
• Both DL-enhanced and standard-of-care scans outperformed unenhanced accelerated scans.
• No significant differences were observed in quantitative volumetric biomarkers or clinical classification between standard-of-care and DL-enhanced accelerated scans.

Conclusions:

• Deep learning reconstruction enables a 60% reduction in MRI sequence scan time.
• High volumetric quantification accuracy and consistent clinical classification are maintained with DL enhancement.
• DL-enhanced accelerated MRI offers superior perceived image quality, supporting its reliability and efficiency for routine clinical use.