Skip to main content
European Radiology Experimental
Fig. 2 | European Radiology Experimental

Fig. 2

From: Artificial intelligence for assessment of vascular involvement and tumor resectability on CT in patients with pancreatic cancer

Fig. 2

The proposed learning framework for PDAC, abdominal organ, and vasculature segmentation on contrast-enhanced CT scans. The teacher segmentation model was trained with 55 manually segmented CT-LAPs of patients with (borderline) resectable and locally advanced PDAC and 50 manually segmented CT-LAPs of control patients. The teacher segmentation model segments the remaining 458 CT-LAPs and CT-PVPs, which together with the 105 manually segmented CT-LAPs, were used to train the student segmentation model. The student segmentation model produced the final segmentations needed to determine vascular involvement and local tumor resectability. CT, Computed tomography; CT-LAP, Late arterial phase CT scan; CT-PVP, Portal venous phase CT scan; PDAC, Pancreatic ductal adenocarcinoma

Back to article page