What question did this study set out to answer?

The aim is to explore AI-based diagnostic strategies for pancreatic ductal adenocarcinoma using CT images. It seeks to identify effective methods for early detection.

March 10, 2026Premier journal of science.0 citationsOpen Access

A Systematic Review on Leveraging Artificial Intelligence for Pancreatic Cancer Diagnosis: Review

SSSonia Suneja

Key Points

The aim is to explore AI-based diagnostic strategies for pancreatic ductal adenocarcinoma using CT images. It seeks to identify effective methods for early detection.
Conducted systematic searches of databases like PubMed, Web of Science, Scopus, and IEEE Xplore.
Implemented PRISMA 2020 guidelines for study identification.
Included studies focused solely on CT imaging techniques for pancreatic tumors.
Categorized studies into four themes regarding AI applications in pancreatic cancer.
A total of 33 studies were included in the final review after screening and applying eligibility criteria.
The studies were categorized into distinct themes, including tumor localization and classification.
AI-driven methods showed potential for enhancing diagnostic accuracy and earlier detection of PDAC.

Abstract

Pancreatic ductal adenocarcinoma (PDAC) remains among the most lethal solid malignancies globally, largely due to delayed diagnosis and limited sensitivity of conventional imaging for early-stage lesions. The substantial mortality burden underscores the urgent need for improved diagnostic strategies capable of identifying subtle radiological patterns in contrast-enhanced computed tomography (CT) scans. In a structured way, all AI-based techniques developed to analyze pancreatic tumors using CT Images from 2018 to 2025 have been evaluated in this review. By conducting systematic searches of various databases, including PubMed, Web of Science, Scopus and IEEE Xplore, the eligible studies were determined. The methods for identifying studies were conducted under the principles of PRISMA 2020. Moreover, the imaging modality used was CT only. The conducted electronic searches of PubMed, Web of Science, Scopus, and IEEE Xplore identified 236 records. Screening was conducted on 195 records, after removal of 41 duplicates. Seventy-eight records were excluded following title and abstract screening. It was sought to retrieve full-text reports for 117 studies; however, 36 could not be retrieved. Out of 81 reports assessed for eligibility, 48 were excluded (20 non-CT imaging modalities; 9 non-pancreatic cancer/PDAC specific; 19 insufficient methodological detail). In total, 33 studies were included. The 33 included studies were systematically categorized into four distinct themes based on their primary focus: (i) AI-driven segmentation for pancreas tumor localization, (ii) deep learning-based tumor classification, (iii) CT-based radiomics and feature-driven analysis, and (iv) early detection models. This review consolidates current advancements in AI-driven frameworks that integrate CT imaging data, thereby enhancing diagnostic accuracy and enabling earlier identification of PDAC. This review aims to bridge this gap by consolidating recent advances in CT-based AI methods for pancreatic disease diagnosis.

Ask AI

Helpful

Bookmark

View Full Paper