Open-source algorithm and software for computed tomography-based virtual pancreatoscopy and other applications

Area of Science:

• Medical Imaging
• Computational Anatomy
• Surgical Technology

Background:

• Pancreatoscopy is crucial for diagnosing pancreatic diseases but carries significant risks due to its invasiveness.
• Virtual pancreatoscopy (VP) presents a less invasive alternative, but its clinical application is limited by low-resolution CT and small pancreatic duct diameters.

Purpose of the Study:

• To develop an open-source virtual pancreatoscopy software platform using an optimal path algorithm and super-resolution technique.
• To enhance the segmentation accuracy and visualization capabilities for the pancreatic duct based on abdominal CT images.

Main Methods:

• Implementation of an optimal path algorithm and super-resolution technique within the 3D Slicer platform.
• Segmentation of the pancreatic duct from abdominal CT images.
• Reconstruction of fly-through visualizations and quantitative analysis of ductal wall thickness and topology.

Main Results:

• Achieved an average Dice coefficient of 0.85 (SD=0.04) for pancreatic duct segmentation, demonstrating high accuracy.
• Successfully reconstructed detailed fly-through visualizations of the pancreatic duct, confirming VP feasibility.
• Enabled quantitative analysis of ductal wall thickness and topology, offering deeper insights into pancreatic diseases.

Conclusions:

• The developed open-source VP system significantly improves the feasibility and clinical potential of virtual pancreatoscopy.
• Advanced segmentation and visualization techniques overcome previous limitations, paving the way for safer pancreatic disease diagnosis and analysis.
• The platform's availability facilitates further research and development in virtual endoscopy for pancreatic conditions.