Journal: NPJ precision oncology
This multicenter study focused on patients with cT1b–T2a clear cell renal cell carcinoma, where some tumors are “occult” pT3a at final pathology—a subgroup with worse prognosis that is difficult to identify preoperatively.
Researchers developed RENALNet, a 3D deep learning model trained on nephrographic-phase CT scans from 1661 patients across five institutions plus the KiTS23 dataset, to predict pathological upstaging to pT3a.
Key points:
- Performance: RENALNet outperformed radiomics-based models in predicting pT3a upstaging.
- Clinical utility: When RENALNet output was combined with assessments from radiologists of different experience levels, overall diagnostic accuracy improved, indicating value as a decision-support tool for surgical planning.
- Interpretability: Grad-CAM heatmaps highlighted anatomically relevant regions of potential invasion, supporting that the model focuses on clinically meaningful features.
- Biological correlation:
- • Higher RENALNet risk scores correlated with higher Ki-67 proliferation indices.
- • RENALNet risk stratified 5‑year progression-free survival, linking the model’s predictions to prognosis.
- Transcriptomics: Tumors classified as high risk by RENALNet showed enrichment of gene expression signatures associated with epithelial–mesenchymal transition, IL6–JAK–STAT3 signaling, and PI3K–Akt signaling—pathways related to tumor aggressiveness.
Overall, RENALNet provides a biologically grounded, interpretable radiogenomic tool that can preoperatively stratify risk of pT3a upstaging in localized clear cell renal cell carcinoma and potentially guide surgical decision-making.