Centerline-guided reinforcement learning model for pancreatic duct identifications.
Sepideh Amiri1, Reza Karimzadeh1, Tomaž Vrtovec2
1University of Copenhagen, Department of Computer Science, Copenhagen, Denmark.
Journal of Medical Imaging (Bellingham, Wash.)
|November 11, 2024
Summary
An automated deep reinforcement learning method accurately traces pancreatic duct centerlines in CT images. This technique shows potential for early pancreatic cancer diagnosis and clinical application.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Pancreatic ductal adenocarcinoma is a leading cause of cancer mortality.
- Early diagnosis is crucial for improving patient outcomes.
- Measuring pancreatic duct diameter is key for early detection.
Purpose of the Study:
- To develop an automated method for tracing pancreatic duct centerlines in CT images.
- To leverage deep reinforcement learning for precise duct measurement.
- To aid in the early diagnosis of pancreatic cancer.
Main Methods:
- A deep reinforcement learning agent interacts with CT images to trace duct centerlines.
- A deep neural network predicts optimal navigational paths.
- Landmark-based registration and gradient methods enhance accuracy.
Main Results:
- The method achieved mean detection errors of 2.0-2.2 mm.
- Hausdorff distances ranged from 4.0-4.9 mm.
- Root mean squared errors were between 2.1-2.6 mm.
Conclusions:
- An automated algorithm for pancreatic duct centerline tracing was developed.
- Validation on external datasets confirms the method's practical utility.
- This approach holds promise for clinical application in early cancer diagnosis.
