Development and validation of a fully automated transformer-based 3D framework for pancreatic fat quantification in pancreatic steatosis

Key Points

• The aim is to develop a fully automated method for quantifying fat in the pancreas using advanced imaging techniques.
• Developed a pipeline for pancreatic segmentation and quantification of fat using MRI data.
• Utilized a transformer-based architecture called TransUNet.
• Combined water-only and fat-only MRI images for accurate results.
• TransUNet achieved the highest accuracy in segmenting pancreatic fat.
• The AI-based method correlated strongly with manual volumetric quantification of fat.
• Outperformed traditional ROI-based quantification methods.