Improved Pancreatic Cancer Detection and Localization on CT Scans: A Computer-Aided Detection Model Utilizing...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Pancreatic ductal adenocarcinoma (PDAC) detection is challenging.
• Early diagnosis significantly impacts patient outcomes.
• Improving CT scan analysis for pancreatic cancer is needed.

Purpose of the Study:

• To develop and evaluate an AI-driven tumor detection framework for pancreatic head cancer.
• To enhance the accuracy of PDAC detection on CT scans.

Main Methods:

• A retrospective study using CT scans from 99 PDAC patients and 98 controls.
• A multi-stage 3D U-Net deep learning model incorporating secondary features (duct dilation).
• Local testing on 59 CT scans and external validation on a public dataset.

Main Results:

• High accuracy in local testing: sensitivity 0.97, specificity 1.00, AUROC 0.99.
• Strong performance in external validation: sensitivity 1.00.
• Acceptable tumor localization accuracy (DSC 0.37).

Conclusions:

• The AI framework effectively detects pancreatic head tumors using CT scans and secondary signs.
• This approach shows high accuracy and potential for improved early diagnosis of PDAC.
• The findings support the integration of AI in radiological assessment for pancreatic cancer.