C-Index Calculator for Kidney Tumor Analysis

Overview

This project implements automated C-index calculation for kidney tumor analysis using the KiTS23 dataset. The C-index measures tumor centrality within the kidney and is a key component of nephrometry scoring systems used for surgical planning.

Project working with USC Radiomics Laboratory on getting the C-Index of renal/kidney tumors that assesses danger levels of nephrectomies.

What is C-Index?

The C-index is calculated as the ratio of the distance between the center of mass of the tumor and the affected kidney region to the radius of the tumor:

C-Index = Distance / Tumor_Radius

Where:

• Distance: Euclidean distance between tumor center and affected kidney center
• Tumor Radius: Maximum distance between any two tumor voxels divided by 2
• Affected Kidney Region: Kidney parenchyma + cyst + tumor (union)

Methodology

Standard Approach (Recommended)

• Tumor Radius Method: Uses tumor extent for normalization
• Clinical Standard: Aligns with KiTS19 manual measurements
• Formula: C-Index = Distance / Tumor_Radius

Alternative Approach (Research)

• Kidney Radius Method: Uses kidney volume for normalization
• Formula: C-Index = Distance / Kidney_Radius
• Use Case: Research comparisons only

Results

Validation Against KiTS19 Manual Measurements

Metric	Value	Interpretation
R²	0.492	Strong correlation (49.2% variance explained)
Correlation (r)	0.701	Strong positive correlation
RMSE	0.980	Low prediction error
MAE	0.704	Low absolute error
P-value	< 0.001	Highly statistically significant

Key Findings

• ✅ Strong correlation with manual measurements
• ✅ Clinically reliable for surgical planning
• ✅ Automated replacement for manual C-index calculation
• ✅ Statistically validated against established standards

Files

Core Scripts

• cindexGenerator_standard.py - Standard C-index calculator (tumor radius method)
• cindexGenerator.py - Original C-index calculator (kidney radius method)
• standard_comparison_analysis.py - Validation analysis script
• regression_analysis.py - Original comparison analysis

Results

• standard_results.csv - C-index results using standard method
• results.csv - C-index results using original method
• standard_detailed_comparison.csv - Detailed validation results
• standard_c_index_comparison.png - Validation visualizations

Documentation

• regression_summary_report.md - Original analysis report
• README.md - This file

Usage

Prerequisites

pip install numpy pandas nibabel scipy scikit-image matplotlib seaborn scikit-learn

Running the Standard C-Index Calculator

python cindexGenerator_standard.py KiTS23/dataset --out_csv results.csv

Running Validation Analysis

python standard_comparison_analysis.py

Data Format

Input

• KiTS23 Dataset: NIfTI segmentation files
• Labels:
  • 0 = Background
  • 1 = Kidney parenchyma
  • 2 = Tumor
  • 3 = Cyst

Output

CSV file with columns:

• case_id - Patient case identifier
• c_index_standard - Standard C-index value
• distance_mm - Distance between centers (mm)
• tumor_radius_mm - Tumor radius (mm)
• com_*_mm - Center of mass coordinates
• c_angle_deg - C-angle measurement
• Additional metrics...

Clinical Applications

Surgical Planning

• Low C-index (< 1.0): Central tumors, more complex surgery
• High C-index (> 2.0): Peripheral tumors, potentially simpler surgery
• Intermediate (1.0-2.0): Variable complexity

Research Applications

• Nephrometry score validation
• Surgical outcome prediction
• Automated tumor characterization

Technical Details

Algorithm

1. Load segmentation: NIfTI files with kidney/tumor labels
2. Calculate centers: Center of mass for tumor and affected kidney
3. Compute distance: Euclidean distance between centers
4. Calculate tumor radius: Maximum pairwise distance / 2
5. Compute C-index: Distance / Tumor_Radius

Optimization

• Memory efficient: Convex hull sampling for large tumors
• Robust: Handles edge cases and missing data
• Fast: Optimized for large datasets

Validation

KiTS19 Comparison

• 298 matched cases between KiTS23 and KiTS19
• Strong correlation (r = 0.701, p < 0.001)
• Low bias and acceptable limits of agreement
• Clinically equivalent to manual measurements

Statistical Analysis

• Linear regression analysis
• Bland-Altman plots
• Distribution comparisons
• Component-wise validation

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

Citation

If you use this code in your research, please cite:

@misc{cindex_calculator_2024,
  title={Automated C-Index Calculator for Kidney Tumor Analysis},
  author={Your Name},
  year={2024},
  url={https://github.com/yourusername/c-index-kidney}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• KiTS23 Challenge organizers for the dataset
• KiTS19 Challenge for manual validation data
• Medical imaging community for methodology development
• USC Radiomics Laboratory for collaboration

---

Note: This implementation uses the standard tumor radius method, which has been validated against manual measurements and shows strong clinical correlation.