Radiomics for early detection of pancreatic cancer: a systematic review and meta-analysis

Background Pancreatic ductal adenocarcinoma (PDAC) is one of the lethal malignancies, in which accurate and faster detection is required in high-risk population to improve prognosis and decrease cancer-associated mortality. Currently, radiomics has emerged as a promising computational approach to address this challenge, reporting increased accuracy in differentiating PDAC from benign lesions. Our study aimed to evaluate radiomics-based models derived from computed tomography, magnetic resonance imaging, positron emission tomography, or ultrasound for the detection of PDAC in patients under surveillance. Methods A systematic literature search on PubMed, Embase, Scopus, and Cochrane was followed by a meta-analysis comparing diagnostic performance metrics, including area under the receiver operating characteristic curve, sensitivity, and specificity. The DerSimonian-Laird method was used to estimate the pooled sensitivity, specificity, positive likelihood ratios (PLRs), and negative likelihood ratios (NLRs), with subgroup analysis performed using Cochrane RevMan 5.4.1 software and OpenMetaAnalyst. Results A total of 15 studies involving 14,688 patients were analyzed, with most studies published between 2019 and 2025. Among these patients, the number of patients with PDAC was 6153 (41.8%), the healthy cases were 7145 (48.6%), and the rest of the patients were unspecified (9.6%). Artificial intelligence (AI)/machine learning (ML) reported a pooled sensitivity of 0.88 (95% CI, 0.84–0.91; I² = 87.8%) and a specificity of 0.93 (95% CI, 0.87–0.96; I² = 95.0%) in detecting PDAC. The pooled PLR was 12.1 (95% CI, 8.4–21.4; I² = 95.5%); however, the NLR was 0.12 (95% CI, 0.09–0.16; I² = 83.1%). Conclusion The use of AI and ML along with diagnostic modality presents a promising alternative to conventional diagnostic modality owing to the display of convincing diagnostic metric for detection of PDAC. Further prospective studies are needed to study the efficacy of this new approach, along with its incorporation with genomic, proteomic, and metabolomic data to develop multi-omic predictive frameworks to further improve PDAC detection.