Work

Peritumoral radiomics indicate a gradient transition between renal tumours and their surrounding healthy tissue: A proof-of-concept analysis

Year: 2025

Type: article

Abstract: PURPOSE: To evaluate whether peritumoral radiomics capture spatial gradients between renal tumors and adjacent healthy tissue, and to assess the ability of tumor-derived radiomic features to classify renal tumor histologic subtypes. METHODS: We analyzed 354 renal tumor cases from the publicly available KiTS23 dataset. Radiomic features were extracted from tumors and concentric peritumoral zones (0-3 mm, 3-6 mm, and 6-9 mm), both within and beyond the kidney. Hierarchical clustering assessed spatial gradients in radiomic signatures. Separately, a multiclass classification model using tumor-only features was developed to differentiate histologic categories. The pipeline included feature selection, and classifier training (logistic regression, random forest, support vector machine). Model interpretability was evaluated with SHAP analysis. RESULTS: Clustering revealed consistent radiomic gradients in peritumoral zones, with patterns varying by histologic subtype. Peritumoral regions often formed clusters distinct from tumor cores, particularly in clear cell and papillary renal cell carcinoma and benign tumors. Chromophobe renal cell carcinoma showed more localized transitions, while other malignancies displayed disrupted gradients. The support vector machineclassifier achieved class-wise ROC AUCs of 0.66-0.78, with balanced performance across histologic groups. CONCLUSION: Peritumoral radiomics demonstrate structured spatial gradients that vary with tumor subtype. These findings highlight the potential of radiomic analysis for probing tumor-tissue interfaces and for supporting the development of non-invasive imaging biomarkers in renal oncology.

Source: European Journal of Radiology

DOI: 10.1016/j.ejrad.2025.112513

Authors: Matthaios Triantafyllou, Savvas Giannettou, Antonios Tzortzakakis, Apostolos H. Karantanas, Michail E. Klontzas

Institutions: University of Crete, University Hospital of Heraklion, Karolinska University Hospital, Karolinska Institutet

Language: English

Topic: Radiomics and Machine Learning in Medical Imaging

Subfield: Radiology, Nuclear Medicine and Imaging

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: hybrid

Estimated APC paid: $3,440