Unmasking pancreatic cancer: Advanced biomedical imaging for its detection in native versus arterial dual‐energy computed tomography (DECT) scans

Abstract

AbstractThis study investigates the potential of a machine learning classifier using dual‐ energy computed tomography (DECT) radiomics to differentiate between malignant pancreatic lesions and normal pancreas tissue. A total of 100 patients who underwent third‐generation DECT between November 2018 and October 2022 were included, with 60 patients having pancreatic cancer and 40 normal pancreatic tissue. Radiomics features were extracted from non‐contrast and arterial‐enhanced DECT scans with stepwise feature reduction used to identify relevant features. Thetrained machine learning classifiers achieved a diagnostic accuracy of 0.97 in the arterial‐enhanced model and 0.88 in non‐contrast scans with sensitivities of 0.97 and 0.96, respectively. Areas under the curve were 0.97 (95% CI, 0.92–1.0, p < 0.001) and 0.96 (95% CI, 0.90–1.0, p < 0.001), respectively with no significant differences between both models (p= 0.52). This approach shows promise in enhancing pancreatic cancer detection and improving patient diagnoses, particulary in specific patient groups.