Three-dimensional texture analysis of renal cell carcinoma cell nuclei for computerized automatic grading

Area of Science:

• Biomedical Engineering
• Computational Pathology
• Digital Image Analysis

Background:

• Accurate grading of renal cell carcinoma (RCC) is crucial for patient treatment.
• Traditional methods often rely on subjective interpretation of cell morphology.
• Objective quantitative analysis of nuclear features can improve diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate a 3D digital image cytometry approach for quantitative analysis of chromatin texture in renal cell carcinoma nuclei.
• To assess the efficacy of 3D texture features for automated cancer cell grading.
• To investigate the combined utility of 3D textural and morphological features for improved classification accuracy.

Main Methods:

• Confocal laser scanning microscopy (CLSM) was used to acquire 3D images of 2,423 cell nuclei from 80 RCC samples.
• 3D texture mapping, 3D gray level co-occurrence matrices, and 3D run length matrices were employed for quantitative texture analysis.
• Discriminant analysis and principal component analysis were utilized for feature selection and classification.

Main Results:

• Automatic grading of cell nuclei using 3D texture features achieved an accuracy of 78.30%.
• Integrating 3D textural features with 3D morphological features enhanced the classification accuracy to 82.19%.
• The study demonstrates the potential of 3D image cytometry for objective RCC grading.

Conclusions:

• Three-dimensional chromatin texture analysis provides valuable quantitative features for renal cell carcinoma grading.
• Combining 3D textural and morphological data significantly improves the accuracy of automated cancer cell classification.
• This approach offers a promising tool for objective and reproducible cancer diagnosis.