Development and Evaluation of Deep Learning-Accelerated Single-Breath-Hold Abdominal HASTE at 3 T Using Variable...

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Imaging Technology
• Artificial Intelligence in Medicine

Background:

• Deep learning (DL) enables faster MRI acquisition but faces challenges with specific absorption rate (SAR) at 3 T.
• Standard T2-weighted fat-suppressed MRI sequences are lengthy and require respiratory triggering.

Purpose of the Study:

• To develop and evaluate a DL-accelerated single-breath-hold HASTE sequence for 3 T abdominal MRI.
• To overcome SAR limitations and compare image quality against the standard BLADE technique.

Main Methods:

• A variable flip angle (FA) evolution for HASTE-DL was optimized in volunteers for low SAR and cardiac motion-related signal loss (CRSL).
• The optimized HASTE-DL sequence was clinically evaluated in 51 patients, comparing it to the T2-weighted BLADE sequence.
• Image quality, artifacts, CRSL, and diagnostic confidence were assessed by two radiologists.

Main Results:

• The optimized HASTE-DL sequence (21 seconds) demonstrated excellent image quality in patients.
• While BLADE was rated higher for image quality parameters, HASTE-DL showed no difference in detecting or measuring hepatic lesions.
• HASTE-DL achieved diagnostic confidence comparable to BLADE despite a significantly shorter acquisition time.

Conclusions:

• A single-breath-hold HASTE-DL sequence using variable refocusing FAs is feasible at 3 T within SAR limits.
• This DL-accelerated technique offers high image quality and diagnostic confidence for abdominal imaging.
• HASTE-DL achieves diagnostic utility at 1/10th the acquisition time of standard T2-weighted sequences.