Related Experiment Video
Updated: Jun 15, 2025
13:56
Robot Assisted Distal Pancreatectomy with Celiac Axis Resection DP-CAR for Pancreatic Cancer: Surgical Planning and Technique
Published on: August 14, 2021
6.3K
A Computer Vision Algorithm to Predict Superior Mesenteric Artery Margin Status for Patients With Pancreatic Ductal
Jane Wang1, Amir Ashraf Ganjouei1, Fernanda Romero-Hernandez1
1Department of Surgery, University of California, San Francisco, CA.
Annals of Surgery
|August 23, 2024
Summary
A new computer vision algorithm can predict superior mesenteric artery (SMA) margin status from CT scans in pancreatic cancer patients. This AI tool shows promise in improving the prediction of vascular invasion during surgery.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in oncology
- Surgical oncology
Background:
- Pancreatic ductal adenocarcinoma (PDAC) requires complete surgical resection for cure.
- Accurate prediction of vascular invasion, particularly of the superior mesenteric artery (SMA), is crucial for successful PDAC resection.
- Current methods for assessing SMA involvement preoperatively have limitations in accuracy.
Purpose of the Study:
- To assess the feasibility of a computer vision algorithm for predicting SMA margin status using preoperative CT scans in PDAC patients.
- To compare the algorithm's predictive performance against expert abdominal radiologists and surgical oncologists.
- To enhance the prediction of vascular involvement in PDAC surgery.
Main Methods:
- Development of a U-Net algorithm for SMA segmentation and a ResNet50 algorithm for margin status prediction from contrast-enhanced CT scans.
- Training and validation using CT scans from 200 adult PDAC patients who underwent the Whipple procedure (2010-2022).
- Blinded review of scans by expert radiologists and surgeons, with pathology reports serving as the reference standard.
Main Results:
- The U-Net model achieved a Dice Similarity Coefficient of 0.90 for SMA segmentation.
- The computer vision algorithm demonstrated the highest sensitivity (0.43) for predicting positive SMA margins compared to radiologists (0.23) and surgeons (0.36).
- Both the algorithm and radiologists achieved excellent specificity (0.94), with the algorithm showing higher overall accuracy (0.85) than radiologists (0.80) and surgeons (0.76).
Conclusions:
- A computer vision algorithm can be feasibly developed to predict SMA margin status from preoperative CT scans in PDAC patients.
- The algorithm shows potential as an adjunct tool to improve the prediction of SMA vascular involvement.
- This AI-driven approach may aid in surgical planning and improve outcomes for PDAC patients.
