Tuesday, December 2 | 3:20 p.m.-3:30 p.m. | SSGI08-3 | Room E353A
For this study, researchers developed an automated framework that may be useful for surgical planning and resection assessment in patients with pancreatic ductal adenocarcinoma (PDAC).
The method quantifies the extent of vascular invasion in PDAC. Called PAN-VIQ, the proposed model integrates multistage annotation and progressive learning to segment tumors and five critical surgical vessels (the celiac artery, common hepatic artery, superior mesenteric artery, superior mesenteric vein, and portal vein).
It then measures vascular invasion via three-dimensional encasement angle analysis, according to presenter Yajiao Zhang, PhD, from Zhejiang University of Traditional Chinese Medicine in Hangzhou, China, and colleagues.
PDAC is the most common and aggressive type of pancreatic cancer. The researchers believe the model can reduce unnecessary exploratory surgeries.
In a multicenter study, Zhang and colleagues conducted a prospective validation evaluating 202 patients treated in 2024. That evaluation revealed "comparable or superior performance" to senior radiologists across various vascular territories, the group found.
Model performance was described as "robust," they noted, highlighting internal validation accuracy of 90.2% (highest performing) for the common hepatic artery, and 87.4% (lowest performing) for the portal vein, among other metrics.
External validation demonstrated accuracies over 90% for all five vessels, with the group reporting the highest accuracy (97.5%) with the celiac artery.
Importantly, PAN-VIQ significantly reduces interobserver variability and outperforms junior radiologists, while maintaining parity with senior experts, the group determined.
Stop by to learn more.
