A practical imaging classification for the non-invasive differentiation of renal cell carcinoma into its main subtypes

Aim Renal cell carcinoma (RCC) is a heterogeneous disease which encompasses various subtypes that exhibit differing biologic behavior and imaging findings. Non-invasive subtype differentiation by imaging facilitates prognostication and treatment selection. The aim of the study was to evaluate the performance of a diagnostic imaging key based on tumor morphology, T2 signal intensity on MRI, and tumor vascularity for differentiating RCC into its subtypes. Materials and methods Using a custom-designed diagnostic imaging key, three blinded fellowship-trained abdominal radiologists independently evaluated the cross-sectional imaging of 50 histologically proven RCCs and categorized these into subtypes in two sessions. The diagnostic performance of the imaging key was evaluated and compared to the baseline performance without the key. Results The 50 RCCs comprised 20 (40%) clear cell, 17 (34%) papillary, and 13 (26%) chromophobe tumors. All expert readers demonstrated an improvement in diagnostic accuracy by an average of 5.3% with the use of the key. The readers showed good to excellent diagnostic performance for clear cell RCC (area under the receiver operating curve, AUROC of 0.86–0.91) and papillary RCC (AUROC of 0.82–0.87), and fair performance with chromophobe RCC (AUROC of 0.67–0.77). The Reader-to-SOR (standard of reference) agreement increased from 0.53 (moderate) to 0.67 (good) with the use of the key. Conclusion The diagnostic imaging key facilitates RCC subtype characterization and can be used as a decision support tool.