Artificial Intelligence in Sincalide-Stimulated Cholescintigraphy: A Pilot Study.

Area of Science:

• Nuclear Medicine
• Artificial Intelligence
• Medical Imaging

Background:

• Sincalide-stimulated cholescintigraphy (SSC) is used to diagnose functional gallbladder disorder by calculating the gallbladder ejection fraction (GBEF).
• AI-driven workflows for real-time image processing and organ function calculation in nuclear medicine are underexplored.

Purpose of the Study:

• To explore an AI-based application for gallbladder radioactivity tracking during SSC.
• To assess the feasibility of AI in real-time gallbladder segmentation and GBEF calculation.

Main Methods:

• Retrospective analysis of 20 SSC exams (10 easy, 10 challenging).
• Manual annotation of gallbladder regions of interest by two human operators.
• Development of a U-Net-based deep learning model for automatic gallbladder segmentation.
• 10-fold cross-validation and comparison of AI-generated masks with human annotations using Dice similarity coefficients (DICE).

Main Results:

• AI achieved an average DICE of 0.746 against operator 1 and 0.676 against operator 2.
• AI performance was better in easy cases (DICE 0.781) than in challenging cases (DICE 0.641).
• Visual inspection revealed AI errors with patient motion or low-count activity.

Conclusions:

• AI shows potential for real-time gallbladder tracking and GBEF calculation in SSC.
• AI-enabled real-time evaluation of nuclear imaging data can advance clinical workflows, offering instantaneous organ function assessments.
• This AI workflow may enhance diagnostic efficiency, reduce scan duration, and improve patient comfort.