Explore RenalTCN
from CT image to segmentation.
An interactive experience for exploring the hybrid Transformer–ConvNet architecture and kidney tumor segmentation.
AI CT & Segmentation Viewer and Simulator
Select a CT slice and run RenalTCN hybrid inference to isolate the kidney and tumor.
Scan Controls
Real-Time Inference Metrics
Processing Time: 18.4 ms
Estimated Tumor Volume: 44.6 cm³
Model Confidence: 99.1%
Tumor Size & Staging Analyzer
Relate the measured maximum diameter to the estimated RenalTCN performance across size categories.
Measured Parameters
Clinical Category: T1b - Medium (4-7 cm)
Tumor confined to the kidney, moderate in size. High model sensitivity.
Estimated Performance by Category
Clinical Stability
Values > 0.90 indicate excellent overlap between the RenalTCN-generated mask and the radiologist’s contour.
Composite Loss Function Optimizer
Simulate the loss-function weights (Focal Loss + Dice Loss + Cross-Entropy) for handling class imbalance.
Training Curve Simulation (Epoch Loss)
Architecture Comparator (RenalTCN vs. State-of-the-Art)
Direct comparison between the efficiency of the hybrid RenalTCN model and standard volumetric networks.
| Model Architecture | Parameters (M) | Kidney Dice Score | Tumor Dice Score | Inference Time (ms/volume) | Status |
|---|---|---|---|---|---|
| RenalTCN (Hybrid) | 18.7M | 98.6% | 94.8% | 18.4 ms | Recommended |
| nnU-Net Standard | 31.2M | 98.1% | 93.5% | 42.1 ms | Standard |
| SwinUNETR 3D | 62.0M | 97.9% | 92.8% | 89.6 ms | High Compute |
| Standard 3D U-Net | 16.5M | 96.5% | 90.1% | 15.2 ms | Fast / Slab |