Comparative study of microwave tomography segmentation techniques based on GMM and KNN in breast cancer detection

Microwave Tomography Imaging (MTI) is a new technology for early breast cancer detection. Compared to other methods such as X-ray, Magnetic Resonance Imaging (MRI) and ultrasound, the MTI technology is almost radiation-free, and low cost. However, the analysis and method to utilize new MTI method still remains unclear. In this paper, we study two segmentation techniques, Gaussian Mixture Model (GMM) and k-Nearest Neighbor (KNN), using the Artificial Neural Network (ANN) tool based on the microwave tomography data, which differentiates normal tissues and suspicious tissues in the breast tissue. Comparing different statistical models in the MTI segmentation process on breast cancer detection, our extensive study contributes to the feature extraction and classification processes on breast cancer detection. The results show that in terms of specificity and Mathew Correlation Coefficient (MCC), the KNN model outperforms the GMM method in segmenting the Region of Interest (ROI) from raw MTI data.