Accelerating Brain MR Imaging With Multisequence and Convolutional Neural Networks.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Magnetic resonance imaging (MRI) is crucial for diagnosis but suffers from long acquisition times.
• Reducing MRI scan duration is essential for improving patient comfort and throughput.
• Deep learning offers potential solutions for accelerating medical image acquisition.

Purpose of the Study:

• To evaluate deep learning techniques for reducing MRI scan time.
• To assess if shared information between MRI sequences can expedite lengthy scans.
• To determine if accelerated MRI maintains diagnostic image quality.

Main Methods:

• Utilized fully sampled T1-FLAIR, T2-FLAIR, and T2WI brain MRI data from 217 patients and 105 healthy subjects.
• Subsampled T1-FLAIR and T2-FLAIR sequences using Cartesian masks at various acceleration factors.
• Employed deep learning reconstruction to predict fully sampled images from undersampled data and T2WI, followed by qualitative and quantitative assessments.

Main Results:

• Radiologists' diagnostic decisions remained consistent between accelerated and fully sampled MRI scans.
• No significant differences were observed in regional signal-to-noise ratios (SNRs) and contrast-to-noise ratios (CNRs) (p > 0.05).
• Qualitative image quality assessments by radiologists showed no significant differences (p > 0.05).

Conclusions:

• Deep learning-based reconstruction effectively accelerates brain MRI acquisition.
• The method produces acceptable image quality without compromising diagnostic accuracy.
• This approach holds promise for significantly reducing patient scan times in clinical practice.