Related Experiment Video
Updated: Sep 14, 2025
An Orthotopic Resectional Mouse Model of Pancreatic Cancer
Published on: September 24, 2020
Utility of a CAD model for estimation of vascular involvement and R0 resection in pancreatic ductal adenocarcinoma
Tianyu Zhao1, Yu Wen2,3,4, Yajie Chen1
1School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan, China.
Background:
The current NCCN guidelines for evaluating vascular involvement in pancreatic ductal adenocarcinoma (PDAC) includes limited evaluation metrics, coarse granularity, and significant inter-observer variability.
Purpose:
This study aimed to develop a computer-aided diagnosis (CAD) model to predict PDAC vascular involvement and R0 Resection by image features extraction.
Materials And Methods:
This study retrospectively included consecutive patients with PDAC after preoperative CT between 2020 and 2022. Based on DoDNet, a segmentation network of PDAC, pancreas and vessels was trained and performed on CT images to extract image features, which characterized morphology, special relationship and texture information. The intraoperative and postoperative histopathological findings were used as reference standards. The assessment of vascular involvement by experienced pancreatic radiologist was the comparison. Features were selected by Fisher Score ranking and Wald Chi-Square Test and merged to predict PDAC vascular involvement.
Results:
A total of 87 patients (mean age: 62.53 ± 7.73 years; 55 males) were evaluated. All patients had histopathologically confirmed PDAC with intraoperative vascular involvement scores, 61 of which had postoperative margin pathology information. A total of 229 image features were extracted. The AUC of CAD model with 3 features (tumor-vessel encasement angle , vascular deformation risk score , tumor morphology risk score ) reached (CA) and partially exceeded (CHA, SMA, PV, SMV) the radiologists' predictions of vascular involvement, where and were significant risk factors respectively for involvement of the 5 major vessels (SMA, CA, CHA, PV, SMV, P < .01) and veins (SMV, PV, P < .05). Specifically, CAD model showed sensitivity comparable to or higher than radiologists, but lower specificity. The CAD reached radiologists' predictions of R0 resection.
Conclusions:
The proposed CAD model extracts CT image features, serving as an objective, reproducible, and accurate tool for predicting PDAC vascular involvement and R0 resection.
