Breast Cancer Detection in Mammogram Images Using K-Means++ Clustering Based on Cuckoo Search Optimization

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Traditional breast cancer detection relies on manual feature extraction from mammograms, which is time-consuming and limited by image quality.
• Existing methods often struggle with feature extraction efficiency and interpretability, impacting diagnostic accuracy.

Purpose of the Study:

• To propose a novel K-means++ clustering based on Cuckoo Search Optimization (KM++CSO) algorithm for automated breast cancer detection.
• To enhance segmentation efficiency and interpretability in mammogram analysis using pre-processing techniques.

Main Methods:

• Implemented a K-means++ clustering algorithm integrated with Cuckoo Search Optimization (KM++CSO).
• Utilized pre-processing techniques, including mathematical morphology and OTSU's thresholding, to improve segmentation.
• Validated the KM++CSO method on three public datasets: Mini-MIAS, DDSM, and BCDR using cross-validation.

Main Results:

• Achieved high detection accuracy: 96.42% (Mini-MIAS), 95.49% (DDSM), and 96.92% (BCDR).
• Attained an average accuracy of 96.27% across all three datasets.
• Obtained a Jaccard index score of 91.05%, indicating strong similarity between detected and reference cancer regions.

Conclusions:

• The proposed KM++CSO method offers a robust and accurate approach for breast cancer detection in mammograms.
• The integration of optimization and pre-processing techniques significantly improves detection performance and interpretability.
• This automated method shows potential for enhancing early breast cancer diagnosis and reducing processing times.