Multistage Segmentation of Prostate Cancer Tissues Using Sample Entropy Texture Analysis.

Area of Science:

• Biomedical image analysis
• Computational pathology
• Digital histopathology

Background:

• Accurate segmentation of cancerous cells in prostate tissue is crucial for diagnosis and treatment.
• Existing texture segmentation methods face challenges in distinguishing benign from malignant regions effectively.

Purpose of the Study:

• To propose a novel multistage segmentation technique for accurate identification of cancerous cells in prostate tissue samples.
• To leverage wavelet packet features and sample entropy for robust texture modeling and gland segmentation.

Main Methods:

• A multistage segmentation approach involving mean-shift algorithm for coarse segmentation.
• Application of wavelet packets for fine gland structure analysis.
• Modeling gland texture using sample entropy to differentiate epithelial and stroma regions.

Main Results:

• The proposed algorithm achieved a 90% classification accuracy in identifying cancerous regions.
• Demonstrated superior performance compared to state-of-the-art texture segmentation techniques based on Dice ratios.
• Fast computation due to efficient modeling with wavelet packet features and sample entropy.

Conclusions:

• The developed multistage segmentation technique offers a robust and accurate method for prostate cancer cell identification.
• The integration of sample entropy features significantly enhances the classification accuracy.
• This approach shows promise for improving diagnostic tools in digital pathology.