Table 3 This table summarizes the models developed for the prediction of patients’ response to treatment, their inputs and performance
From: Navigating advanced renal cell carcinoma in the era of artificial intelligence
| Â | Model | Input | Performance/Findings | Limitations |
|---|---|---|---|---|
Rossi et al. | CT-based radiomic analysis | Radiomic features from CT scans of 53 mRCC patients | Correlated radiomic features with progression as the best response to ICI therapy | - Small sample size -Retrospective study - No validation cohorts |
Park et al. | Clinical-CT texture models | Baseline and follow-up CT texture data combined with clinical data in 129 patients | Combined model predicted overall survival (C-index 0.7) and progression-free survival (C-index 0.63), outperforming clinical data alone | - Small sample size -Retrospective study - No validation cohorts -ROI determined by only one radiologist |
Khene et al. | Texture analysis for survival prediction | Pre-treatment CT texture features in 48 patients | Identified predictors of overall and progression-free survival in mRCC patients treated with nivolumab. | - Small sample size -Retrospective study - No validation cohorts -Manual ROIs -Variable CT techniques |
Neutrophil to Lymphocyte Ratio Studies | Biomarker analysis | Neutrophil to lymphocyte ratio | Low splenic volume 3 months after ICI treatment linked to improved overall survival | -Retrospective study - No validation cohorts |
Splenic Volume Studies | Automated splenic segmentation | Splenic volume change measured using AI tools | Significant survival improvement associated with low splenic volume | - Small cohorts - Mixed treatment regimens - No validation cohorts |
Negreros-Osuna et al. | CT-based radiomic model for TKI response prediction | Texture analysis of primary tumors and clinical data in 62 patients | Combined radiomic and clinical model outperformed models using radiomic or clinical data alone. | - Small sample size -Retrospective study - No validation cohorts |
Chen et al. | CT-based radiomic model for short-term lesion response prediction | Radiomic features from baseline arterial phase (AP) and non-contrast (NC) CT scans in 36 patients with recurrent or mRCC | Delta feature-based model effectively predicted short-term lesion response to first-line TKIs in a small cohort. 0.940 (95% CI, 0.890‒0.990) in the training cohort and 0.916 (95% CI, 0.828‒1.000) in the validation cohort | -Small sample size -Retrospective study |
Udayakumar et al. | Radiogenomic model using DCE-MRI | DCE-MRI, histopathology, and transcriptome correlatives in 49 ccRCC patients | High arterial spin labeling MRI correlated with favorable response to antiangiogenic regimens | -Small sample size - Cohort included mostly small tumors -Colocalization between presurgical imaging and postsurgical histologic analysis |