Prospective and multi-reader evaluation of deep learning reconstruction-based accelerated rectal MRI: image quality,...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Oncology Imaging

Background:

• Rectal cancer diagnosis relies heavily on Magnetic Resonance Imaging (MRI) for accurate staging.
• Standard MRI protocols can be time-consuming, potentially impacting patient throughput and comfort.
• Deep learning reconstruction (DLR) offers a potential solution for accelerating MRI acquisition without compromising image quality.

Purpose of the Study:

• To compare the efficacy of DLR-based accelerated rectal MRI (FSE_DL) against standard MRI (FSE_standard).
• To evaluate image quality, diagnostic performance, and reading times for staging rectal adenocarcinoma.

Main Methods:

• Prospective intra-individual comparison of FSE_standard and FSE_DL sequences in 117 patients with rectal adenocarcinoma.
• Quantitative and qualitative image quality assessment, including signal-to-noise and contrast-to-noise ratios.
• Analysis of diagnostic performance for T-staging, N-staging, extramural vascular invasion (EMVI), and mesorectal fascia (MRF) status, using histopathology as the gold standard.

Main Results:

• FSE_DL reduced acquisition time by 65% compared to FSE_standard.
• FSE_DL demonstrated superior image quality, including higher SNR, CNR, and improved subjective scores for clarity and confidence (p < 0.001).
• Junior radiologists showed significantly improved T-staging accuracy and reduced reading times with FSE_DL (p < 0.05).

Conclusions:

• DLR-based accelerated rectal MRI (FSE_DL) is clinically applicable, offering enhanced image quality and a substantial reduction in scanning time.
• FSE_DL improves diagnostic efficiency, particularly for junior radiologists, by enhancing T-staging accuracy and reducing evaluation time.
• The DLR technique shows promise in optimizing rectal MRI examinations, potentially easing the burden on patients and healthcare systems.