The accurate segmentation of pancreatic subregions (head, body, and tail) in CT images provides an opportunity to examine the local morphological and textural changes in the pancreas. Quantifying such changes aids in understanding the spatial heterogeneity of the pancreas and assists in the diagnosis and treatment planning of pancreatic cancer. Manual outlining of pancreatic subregions is tedious, time-consuming, and prone to subjective inconsistency. This paper presents a multistage anatomy-guided framework for accurate and automatic 3D segmentation of pancreatic subregions in CT images. Using the delineated pancreas, two soft-label maps were estimated for subregional segmentation—one by training a fully supervised naïve Bayes model that considers the length and volumetric proportions of each subregional structure based on their anatomical arrangement, and the other by using the conventional deep learning U-Net architecture for 3D segmentation. The U-Net model then estimates the joint probability of the two maps and performs optimal segmentation of subregions. Model performance was assessed using three datasets of contrast-enhanced abdominal CT scans: one public NIH dataset of the healthy pancreas, and two datasets D1 and D2 (one for each of pre-cancerous and cancerous pancreas). The model demonstrated excellent performance during the multifold cross-validation using the NIH dataset, and external validation using D1 and D2. To the best of our knowledge, this is the first automated model for the segmentation of pancreatic subregions in CT images. A dataset consisting of reference anatomical labels for subregions in all images of the NIH dataset is also established.
Background/Objectives: Pancreatic ductal adenocarcinoma (PDAC) prediction in high-risk individuals is essential for early detection and improved outcome. While prior studies have utilized pancreatic radiomics for PDAC prediction, the added value of main pancreatic duct (MPD) features remains unclear. This study aims to assess the additional value of features of the main pancreatic duct (MPD) for predicting PDAC occurrence across different timeframes in advance. Methods: In total, 321 contrast-enhanced CT scans of the MPD and pancreas carried out across control, pre-diagnostic, and diagnostic cohorts were segmented, and radiomics were extracted. A support vector machine (SVM) classifier was used to classify the control and pre-diagnostic cohorts, with model performance assessed using area under the receiver operating characteristic (ROC) curves (AUCs) Results: The MPD diameter and volume significantly increased from the control to the pre-diagnostic and diagnostic CT scans (p < 0.05). The addition of features of the MPD to the pancreas improved the PDAC prediction AUC from 0.83 to 0.96 for subjects 6 months to 3 years in advance, from 0.81 to 0.94 for 3–6 years in advance, and 0.75 to 0.84 for 6–10 years in advance of diagnosis. Additionally, integrating MPD radiomics with diameter and volume significantly improved the AUC from 0.81 to 0.88 for subjects 6 months to 3 years in advance. Conclusions: Radiomic features from abdominal CT scans allow PDAC prediction up to 10 years in advance. Integrating MPD features, including diameter and volume, significantly improves PDAC prediction compared to using radiomics of the pancreas alone.