A fully automated pancreatic subregion segmentation tool for CT

CT imaging is widely used to analyze pancreatic pathologies as it provides useful information regarding progression of pancreatic disease. Segmentation of the pancreas is a pre-requisite for many applications, such as longitudinal change tracking. While current approaches segment the whole pancreas and extract CT biomarkers from the segmentation, changes in pancreatic morphology and CT attenuation in certain pancreatic sub-regions may provide more details related to disease severity and prognosis. In this work, an automated 3D tool was developed to segment three main pancreatic sub-regions (the head, body, and tail) on CT, thereby allowing biomarkers to be extracted from each sub-region. Then, the proposed tool was evaluated internally and externally on two large public datasets: TotalSegmentator and PANORAMA. On the TotalSegmentator dataset, the mean Dice similarity coefficient (DSC, %) and normalized surface distance (NSD, %) for the head, body, and tail were 89.3 ± 13.5 (std. dev.) and 93.9 ± 12.4, 84.9 ± 11.5 and 89.7 ± 10.2, and 84.5 ± 13.9 and 89.9 ± 11.9, respectively. On the PANORAMA dataset, the DSC and NSD for the head, body, and tail were 84.5 ± 7.7 and 87.1 ± 9.5, 81.7 ± 8.9 and 87.8 ± 9.2, and 80.7 ± 14.9 and 88.4 ± 14.3, respectively. The proposed tool can accurately segment pancreatic sub-regions (head, body, and tail), such that biomarkers can be extracted to correlate against morphological and functional changes due to disease.