Deep Learning HASTE for Upper Abdominal MRI: Improved Image Quality, Speed, and Energy Efficiency in a Prospective...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Conventional T2-weighted Half-Fourier Acquisition Single-shot Turbo spin Echo (HASTE) sequences are standard for upper abdominal MRI.
• Deep learning (DL) techniques are emerging to enhance MRI acquisition efficiency and image quality.

Purpose of the Study:

• To qualitatively and quantitatively compare DL-reconstructed T2-weighted HASTE sequences with conventional T2-weighted HASTE sequences for 3T upper abdominal MRI.
• To assess image quality, anatomical visualization, diagnostic performance, and efficiency metrics.

Main Methods:

• Prospective enrollment of 166 patients undergoing 3T upper abdominal MRI.
• Comparison of conventional T2-weighted HASTE MRI with DL-reconstructed fast T2-weighted HASTE MRI.
• Image quality and diagnostic performance assessed by three readers using Likert scales.
• Quantitative analysis of signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and radiomic features.

Main Results:

• DL HASTE demonstrated significantly superior image quality, fewer artifacts, and improved anatomical structure delineation (p < 0.001).
• DL HASTE showed significantly better SNR (p < 0.001) but comparable CNR.
• Significant time reduction (62.5%) and energy cost savings (0.34 kW/scan) were observed with DL HASTE.

Conclusions:

• DL-reconstructed HASTE sequences offer enhanced image quality, reduced artifacts, and improved diagnostic confidence for upper abdominal MRI.
• DL HASTE significantly decreases acquisition time and energy consumption, presenting a more efficient and potentially impactful alternative to conventional methods.
• The findings suggest DL HASTE has the potential for more accurate pathology detection in clinical practice.