MRI sequence focused on pancreatic morphology evaluation: three-shot turbo spin-echo with deep learning-based...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Early detection of pancreatic cancer requires higher-resolution magnetic resonance imaging (MRI) sequences.
• Conventional MRI sequences like HASTE, FS-3D-T1W, and MRCP have limitations in pancreatic imaging quality.

Purpose of the Study:

• To compare a novel T2-weighted, high-contrast, thin-slice MRI sequence with deep learning reconstruction (3S-TSE-DLR) against conventional sequences.
• To evaluate the image quality, including artifact reduction and margin sharpness, of 3S-TSE-DLR for pancreatic imaging.

Main Methods:

• Fifty healthy volunteers underwent pancreatic MRI using 3S-TSE-DLR, HASTE, FS-3D-T1W, and MRCP.
• Two radiologists assessed motion artifacts, pancreatic margin sharpness, and main pancreatic duct (MPD) identification using a 5-point scale.

Main Results:

• 3S-TSE-DLR demonstrated significantly higher artifact scores than HASTE and significantly lower scores than FS-3D-T1W.
• Image sharpness was significantly better with 3S-TSE-DLR compared to both HASTE and FS-3D-T1W.
• MPD identification rates were comparable between 3S-TSE-DLR and MRCP, and superior to FS-3D-T1W and HASTE for at least one radiologist.

Conclusions:

• The 3S-TSE-DLR sequence provides enhanced image sharpness for pancreatic imaging.
• This novel sequence offers comparable or superior visualization of the main pancreatic duct compared to conventional MRI techniques.
• 3S-TSE-DLR shows potential for improving early detection of pancreatic pathologies.