This is an early access version, the complete PDF, HTML, and XML versions will be available soon.
CT-Based Radiomics in the Characterization of Solid Renal Tumors: A Systematic Review
by
, , , and
Petros Koumpis
1
,
Eyrysthenis Vartholomatos
1,
Eleni Romeo
2
,
George A. Alexiou
2
,
Maria I. Argyropoulou
3 and
Athina C. Tsili
3,* 1
Department of Clinical Radiology, University Hospital of Ioannina, University Campus, 45110 Ioannina, Greece
2
Department of Neurosurgery, Faculty of Medicine, School of Health Sciences, University of Ioannina, University Campus, 45110 Ioannina, Greece
3
Department of Clinical Radiology, Faculty of Medicine, School of Health Sciences, University of Ioannina, University Campus, 45110 Ioannina, Greece
*
Author to whom correspondence should be addressed.
Cancers 2026, 18(11), 1758; https://doi.org/10.3390/cancers18111758
Submission received: 7 April 2026
/
Revised: 20 May 2026
/
Accepted: 24 May 2026
/
Published: 27 May 2026
(This article belongs to the Section Methods and Technologies Development)
Simple Summary
Renal cell carcinoma (RCC) presents a significant diagnostic challenge due to its marked heterogeneity and the high prevalence of benign tumors, such as fat-poor angiomyolipoma (fpAML) and renal oncocytoma (RO), which are frequently overtreated surgically. This systematic review, encompassing 47 studies and 11,999 patients, evaluates the efficacy of CT-based radiomics as a “virtual biopsy” tool for the non-invasive characterization of solid renal tumors. The analysis demonstrates high diagnostic accuracy, with a median Area Under the Curve of 0.830 (0.747–0.900) for differentiating benign tumors from RCC, 0.900 (0.861–0.910) for clear cell RCC vs. non-clear cell RCC discrimination, 0.912 (0.879–0.933) for fpAML vs. RCC identification, and 0.885 (0.841–0.947) for RO vs. RCC differentiation. Notably, combined nomograms provide the most accurate predictions, although the number of qualifying studies remains small. Ultimately, CT radiomics offers a non-invasive method for repeated evaluation of intratumoral heterogeneity, with potential applications in personalized treatment strategies.
Abstract
Background: Renal cell carcinoma (RCC) is a global health challenge characterized by significant histological heterogeneity. Conventional contrast-enhanced CT often struggles to differentiate RCC from solid benign renal tumors like fat-poor angiomyolipoma (fpAML) and renal oncocytoma (RO), leading to potential surgical overtreatment. CT-based radiomics has emerged as a promising non-invasive approach that extracts high-dimensional quantitative imaging features to support lesion characterization and may contribute toward more comprehensive, biopsy-adjacent decision support, although it does not yet replace histopathological assessment. Methods: This review systematically evaluates the predictive performance of CT-based radiomics in characterizing solid renal tumors. A literature search was conducted in PubMed/MEDLINE, Cochrane, and Scopus databases for original research published between 2012 and 2025. The review focuses on four key areas: differentiating benign renal tumors from RCC, clear cell (ccRCC) from non-ccRCC, fpAML from RCC, and RO from RCC. Results: In total, 47 studies were assessed, including 11,999 patients. CT-based radiomics demonstrates high diagnostic performance across all categories. Median Area Under the Curve values were 0.830 (0.747–0.900) for benign vs. malignant differentiation, 0.900 (0.861–0.910) for ccRCC vs. non-ccRCC, 0.912 (0.879–0.933) for fpAML vs. RCC, and 0.885 (0.841–0.947) for RO vs. RCC. The integration of radiomic features with clinical parameters into combined nomograms consistently yielded the highest predictive accuracy. Conclusions: Radiomics provides a non-invasive, objective method to characterize renal tumors, potentially reducing unnecessary surgeries and enabling personalized treatment. However, widespread clinical adoption remains limited by a lack of protocol standardization, the need for automated segmentation, and the requirement for prospective, multicenter validation.
