Associations between CT radiomics analyses and kidney biopsy in patients with kidney disease

Kidney disease is characterized by microstructural alterations that currently require invasive biopsy for definitive assessment. However, it remains unclear to what extent radiomics features extracted from contrast-enhanced CT can non-invasively reflect kidney function and histopathological changes.

Between October 2020 and May 2025 all patients undergoing kidney biopsies and having CT scans prior to biopsy were retrospectively analyzed. A total of 49 patients (59% female, median age 60 years) were included. Of the included patients, 35 biopsies were performed in native kidneys (71%) and 14 in kidney allografts (29%). Contrast-enhanced CT images were used to extract radiomics parameters of the kidney. Kidney segmentation was performed using TotalSegmentator and radiomics feature extraction was conducted with PyRadiomics.

Several associations were identified between the extracted radiomics features and kidney function as well as kidney tissue alterations. For the eGFR (CKD-EPI) the highest association was found for the radiomics feature Energy, which is a measurement of the intensity uniformity (ρ = 0.51, p < 0.001), while the first-order feature 90th Percentile showed the best performance in discriminating patients above and below an eGFR threshold of 15 mL/min/1.73 m² with an AUC of 0.83 (95% CI: 0.67-0.98, p = 0.001). Busyness correlated negatively with glomerulosclerosis (ρ = -0.38, p = 0.007), and Coarseness was positively associated with interstitial inflammation (ρ = 0.37, p = 0.008).

CT radiomics features are associated with kidney function as well as histopathological alterations observed in kidney biopsies. Further validation is needed in future studies.