Combination of CT-based Radiomics Features and Clinical Data for Predicting Tumor Genetic Profile in patients with Intrahepatic Cholangiocarcinoma

Background: Intrahepatic cholangiocarcinoma (iCCA) is an aggressive cancer with increasing incidence, and its genetic alterations may serve as potential targets for systemic therapies. Objective: This study aims to determine whether radiomic features extracted from contrast-enhanced CT scans can predict iCCA genetic alterations non-invasively. Methods: Consecutive patients diagnosed with mass-forming iCCA between January 2016 and June 2022 were included. Criteria for inclusion were high-quality CT scans and molecular profiling (NGS or FISH for FGFR2 fusion/rearrangement). Genetic analysis was conducted on surgical specimens (for resectable patients) or biopsies (for unresectable ones). Radiomic features were extracted using LifeX software. Multivariate models were developed to predict common genetic alterations. Results: The study included 90 patients (58 with NGS, 32 with FISH, median age 65). The most common genetic alterations were FGFR2 (20/90), IDH1 (10/58), and KRAS (9/58). Combined clinical-radiomic models showed the best performance in predicting FGFR2 (AUC = 0.892) and IDH1 status (AUC = 0.819), outperforming clinical or radiomic models alone. The radiomic model for KRAS mutations achieved an AUC of 0.767, compared to 0.660 for the clinical model, with no significant improvement when clinical features were added. Conclusion: CT-based radiomics offers a reliable, non-invasive method for predicting genetic alterations in iCCA, with potential implications for personalized treatment strategies.

Grantor dc:publisher

Università degli Studi di Cagliari

Year dc:date

2025

Author dc:creator

• LAINO, MARIA ELENA

Contributors dc:contributor

• SABA, LUCA

• Settore MED/36 - Diagnostica per Immagini e Radioterapia

Statement dc:rights

• info:eu-repo/semantics/openAccess

Language dc:language

ita

Handle dc:identifier

https://hdl.handle.net/11584/447169

OAI identifier oai:identifier

oai:iris.unica.it:11584/447169