BackgroundDetermination of vessel resection in patients with pancreatectomy after neo-adjuvant chemotherapy remains controversial. The recently introduced computed tomography-based vascular burden index presents a potential solution to this challenge. This study aimed to evaluate the model performance for the prediction of vascular resection and pathological invasion. MethodsPatients who underwent surgery after neo-adjuvant chemotherapy were included. Two independent reviewers measured the vascular tumour burden index around the adjacent artery (AVBI), and vein (VVBI). The area under the curve was compared to assess the predictive capacity of vascular burden index values and their changes for vascular resection and pathological vascular invasion. ResultsAmong 252 patients, 179 and 73 had borderline resectable and locally advanced pancreatic cancer, respectively. Concurrent vessel resection and pathological vascular invasion were observed in 121 (48.0 %) and 42 (16.6 %) patients, respectively. In all patients, the VVBI (area under the curve: 0.872) and AVBI (0.911) after neo-adjuvant therapy significantly predicted vessel resection. In patients with vascular resection, the VVBI after neo-adjuvant chemotherapy (0.752) and delta value of the AVBI (0.706) demonstrated better performance for predicting pathological invasion of the resected vein. The regression of the AVBI and VVBI was an independent prognostic factor for survival (hazard ratio: 0.54, 95 % confidence interval: 0.34–0.85; P = 0.009) ConclusionsRegressed VVBI on serial computed tomography scans is useful for predicting vein resection and pathological venous invasion before surgery. The delta value of the AVBI may therefore be helpful for predicting pathological arterial invasion after neo-adjuvant chemotherapy.