Deep learning reconstruction in MRCP: Impact on IPMN characterization and common bile duct stone detection

Area of Science:

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

Background:

• Magnetic Resonance Cholangiopancreatography (MRCP) is crucial for diagnosing ductal pathologies like Intraductal Papillary Mucinous Neoplasms (IPMN) and common bile duct (CBD) stones.
• Image quality limitations in MRCP, including noise and artifacts, can impact diagnostic accuracy.
• Deep learning (DL) reconstruction offers a potential solution to enhance MRCP image quality and diagnostic confidence.

Purpose of the Study:

• To compare the diagnostic confidence of DL versus non-DL reconstructed MRCP images.
• To evaluate the effectiveness of DL in improving image quality for IPMN and CBD stone detection.

Main Methods:

• A retrospective study of 91 patients undergoing MRCP for IPMN or CBD stone evaluation.
• Analysis of 2D and 3D MRCP sequences with and without DL reconstruction.
• Objective (SNR, CR, CNR) and subjective (image quality, artifacts, noise, contrast) assessments by two radiologists, including diagnostic confidence scores.

Main Results:

• DL reconstruction significantly improved SNR, contrast, and CNR in the CBD, especially in 2D sequences.
• Subjective analysis showed DL images had higher ratings for contrast, reduced noise, and better overall quality.
• Diagnostic confidence was enhanced for IPMN typing, malignancy assessment, and CBD stone detection.

Conclusions:

• DL reconstruction demonstrably improves MRCP image quality and reader diagnostic confidence for IPMN and CBD stone evaluation.
• Integration of DL technology into clinical MRCP protocols is supported by these findings.
• Further prospective studies are warranted to assess the impact on diagnostic accuracy and patient outcomes.