Dual-Energy CT (DECT) and Radiomics for the Assessment of Treatment Response in Peritoneal Metastases Undergoing Locoregional Therapy.

Background: Peritoneal metastases (PM) are associated with poor prognosis, and increasing use of locoregional treatments such as Pressurized Intraperitoneal Aerosol Chemotherapy (PIPAC) requires reliable tools for response assessment. Conventional CT evaluation is limited by post-treatment fibrotic changes and the inadequacy of RECIST criteria for small peritoneal lesions, highlighting the need for predictive imaging biomarkers beyond radiological PCI reduction. Aim: This thesis project aimed to identify quantitative features derived from dual energy CT and radiomics predictive of response to locoregional therapy in patients with PM undergoing PIPAC. Methods: Between 2020 and 2025, 37 patients with PM (73 nodules) who underwent pre- and post-treatment CT were retrospectively analysed. Response was defined according to RECIST, surgical PCI variation, and eligibility for cytoreductive surgery (22 responders, 15 non-responders). Lesions were segmented on pre-treatment CT, and 843 radiomic features were extracted. Dual-energy CT parameters were analysed in a subgroup of 15 patients. Results: Spectral parameters showed significant differences between responders and non-responders (p < 0.05). Responders demonstrated a greater reduction in normalised iodine concentration (p = 0.04). Reproducibility assessment of radiomics features demonstrated an excellent intrareader agreement (mean ICC 0.98, 95%IC 0.97-0.99) and a good interreader agreement (mean ICC 0.68, 95%IC 0.63-0.73). LASSO regression identified two clinical variables (increase in radiological PCI and ascites) and two texture-based radiomic features (GLCM Difference Entropy and GLRLM Short Run Low Grey Level Emphasis) associated with response. The combined model achieved a mean AUC of 0.75 (95% CI 0.62–0.84). Conclusion: Although limited by sample size and lack of external validation, integrated spectral and radiomic analysis may improve the prediction of therapeutic response in PM, supporting more accurate patient stratification toward precision medicine strategies.