Radiologic and Lipid Metabolism Imaging Features Associated with FABP2 Expression in Clear Cell Renal Cell Carcinoma: An Interpretable Supervised Machine Learning Radiogenomic Study.
Authors
Affiliations (6)
Affiliations (6)
- Ultrasound Radiogenomics AI Center, 72026 San Pancrazio Salentino, Italy.
- Department of Radiology, Cittadella della Salute, Azienda Sanitaria Locale di Lecce, 73100 Lecce, Italy.
- Department of Medicine and Surgery, Research Unit of Radiology, Campus Bio-Medico University of Rome, 00128 Rome, Italy.
- Apphia Srl, 73100 Lecce, Italy.
- Department of Urology, Vito Fazzi Hospital, 73100 Lecce, Italy.
- Fondazione Policlinico Universitario Campus Bio-Medico, 00128 Rome, Italy.
Abstract
This study investigated the association between computed tomography (CT) features and fatty acid-binding protein 2 (FABP2) gene expression in clear cell renal cell carcinoma (ccRCC), focusing on morphologic and metabolic parameters. A retrospective analysis was performed on 246 ccRCC patients from The Cancer Genome Atlas (TCGA) and The Cancer Imaging Archive (TCIA). Patients were classified as FABP2-positive (n = 57, 23.17%) or FABP2-negative (n = 189, 76.83%). Clinical, pathological, and imaging features were compared. Quantitative CT assessed mean tumor attenuation and abdominal adipose compartments. Two logistic regression models were built: one based on radiologic features and one on metabolic imaging parameters. FABP2-positive tumors were significantly smaller, more frequently low-grade and diagnosed at an earlier stage. FABP2 expression was more common in male patients. On imaging, FABP2-negative tumors more often exhibited collateral vascular supply. Quantitative CT revealed lower mean attenuation in FABP2-positive tumors, suggesting higher intracellular lipid content. No significant differences were found in hepatic attenuation or abdominal adipose compartments. The radiologic model achieved an area under the curve (AUC) of 0.58, with cystic composition as the strongest predictor of FABP2 positivity. The lipid metabolism-based model showed slightly better performance (AUC 0.61), with mean tumor attenuation as the most influential variable. FABP2-positive ccRCC appears to represent a less aggressive, lipid-rich phenotype. CT attenuation features were weakly associated with FABP2 expression and may offer preliminary, noninvasive insights to support integrated radiogenomic approaches in ccRCC.