Breast Cancer Detection Using Automated Segmentation and Genetic Algorithms.

Area of Science:

• Oncology
• Medical Imaging
• Bioinformatics

Background:

• Breast cancer is a leading cause of death in women globally.
• Early detection significantly improves patient outcomes and prevents metastasis.
• Accurate classification of benign and malignant lesions is crucial for effective treatment.

Purpose of the Study:

• To develop an advanced bio-marker for classifying breast lesions as benign or malignant.
• To integrate Bayesian predictive models, pyRadiomics, and genetic algorithms for enhanced diagnostic accuracy.
• To evaluate the effectiveness of automated whole-breast analysis compared to radiologist-segmented lesions.

Main Methods:

• A novel bio-marker was designed using Bayesian predictive models, pyRadiomics System, and genetic algorithms.
• The method analyzed two types of image data: radiologist-segmented lesions and automated whole-breast analysis.
• Performance was evaluated by comparing classification effectiveness and diagnostic accuracy (AUC).

Main Results:

• The automated whole-breast analysis showed a minor 12% effectiveness difference for calcifications compared to radiologist segmentation.
• A 25% effectiveness difference was observed for masses between lesion-specific and whole-breast analyses.
• The approach achieved high diagnostic accuracy with an AUC of 0.86 for calcifications and 0.96 for masses.

Conclusions:

• The proposed bio-marker effectively classifies benign and malignant breast lesions.
• Automated whole-breast analysis offers a viable alternative to radiologist segmentation, with comparable effectiveness for calcifications.
• This method demonstrates significant potential for improving early breast cancer detection and diagnosis.