A Two-Phase Approach using Mask R-CNN and 3D U-Net for High-Accuracy Automatic Segmentation of Pancreas in CT Imaging

Ramazan Özgür Dogan, Hülya Dogan, Coskun Bayrak, Temel Kayikçioglu

Abstract: Highlights•Automatic pancreas segmentation techniques decrease the rate of pancreatic disease deaths and assist the medical doctor in the analysis of pancreatic diseases.•Owing to more complex anatomical structures (size, shape and position) of the pancreas, pancreas segmentation techniques cannot provide satisfactory performance and require significantly higher computational power and memory.•A novel two-phase approach for high-accuracy automatic pancreas segmentation in computed tomography (CT) imaging is suggested in this study.•The determination of the rough position of the pancreas with different anatomical structures (Pancreas Localization phase) is a critical phase for automatic pancreas segmentation with more satisfactory performance.