Value of radiomics features extracted from baseline computed tomography images in predicting overall survival in patients with nonsurgical pancreatic ductal adenocarcinoma: incorporation of a radiomics score to a multiparametric nomogram to predict 1-year overall survival.

Purpose: This study aimed to determine the value of radiomics features derived from baseline computed tomography (CT) scans and volumetric measurements to predict overall survival (OS) in patients with nonsurgical pancreatic ductal adenocarcinoma (PDAC) treated with a chemotherapy combination regimen of 5-fluorouracil, leucovorin, irinotecan, and oxaliplatin (FOLFIRINOX).
Methods: In this retrospective single-institution study, 131 patients with nonsurgical PDAC who received FOLFIRINOX neoadjuvant chemotherapy between December 2012 and November 2021 were included. Pretreatment contrast-enhanced CT images were obtained for all patients before inclusion. The primary tumor was contoured by an expert radiologist with 25 years of experience. A total of 845 radiomics features, including first-, second-, and higher-order features, were extracted from the total tumor volume. A feature reduction pipeline was used to reduce the dimensionality of the data. The selected features were used to generate a radiomics score based on the Least Absolute Shrinkage and Selection Operator coefficients. A high-dimensional Cox model was generated on the basis of the radiomics score and other quantitative and semantic imaging findings.
Results: From the 845 radiomics features extracted, 45 were significantly different between the tertiles. The following equation was used to generate a radiomics score: radiomics score = SmallAreaEmphasis (-66.87801 + LargeDependenceEmphasis) - 0.2345916. The radiomics score was significantly different among the 3 groups of the radiomics features (P = .034). The overall difference in survival was significant among the 3 groups (P = .02). The nomogram showed good calibration and showed significant differences among the patients when they were classified as tertiles (P < .00).
Conclusion: Radiomics approaches have the potential to predict OS in nonsurgical patients with PDAC, and the inclusion of semantic imaging findings and pathologic data could further enhance prognostication in patients with PDAC.
Competing Interests: Declaration of competing interest The authors declare no competing interests.
(Copyright © 2024. Published by Elsevier Inc.)