Work

A multimodal MRI radiomics sonata for differentiating small renal masses: a two-center retrospective study

Year: 2026

Type: article

Abstract: BACKGROUND: Accurately distinguishing benign from malignant small renal masses (SRMs) is crucial to avoid unnecessary surgeries, yet reliable preoperative diagnostic tools remain limited. METHODS: A total of 189 patients with SRMs from two hospitals (2018-2022) were enrolled. Clinical data and MRI features were analyzed by two radiologists. Radiomics features were automatically extracted from manually segmented ROIs using dedicated software. Significant features were selected via univariate analysis, multivariate regression, and Lasso to develop three models: radiomics, clinical, and combined nomogram. Diagnostic performance was evaluated using the Area Under the receiver operating characteristic Curve (AUC). RESULTS: The radiomics-clinical nomogram was constructed using T2WI signal intensity (p < 0.001), pseudocapsule (p = 0.001), and radiomics score (Rad-score, p = 0.001) via multivariate logistic regression. In the training set, the nomogram achieved an AUC of 0.962 (95% CI: 0.931-0.993), outperforming both the radiomics model (AUC = 0.918, 95% CI: 0.862-0.974) and the clinical model (AUC = 0.856, 95% CI: 0.780-0.933). Favorable performance was observed in the internal validation set (nomogram: AUC = 0.981, 95% CI: 0.944-1.000; radiomics model: AUC = 0.912, 95% CI: 0.750-1.000; clinical model: AUC = 0.844, 95% CI: 0.578-1.000) and external validation set (nomogram: AUC = 0.971, 95% CI: 0.886-1.000; radiomics model: AUC = 0.928, 95% CI: 0.841-0.989; clinical model: AUC = 0.916,95% CI: 0.760-0.976). CONCLUSION: All three models enable effective and non-invasive preoperative differentiation of benign from malignant SRMs, with the radiomics-clinical nomogram exhibiting the highest diagnostic accuracy, supporting its potential as a valuable clinical tool.

Source: Clinics

DOI: 10.1016/j.clinsp.2026.101212