Automated Breast Tumor Detection and Segmentation Using the Threshold Density Algorithm with Logistic Regression on Microwave Images

Context: Breast cancer remains a major health burden worldwide, necessitating improved screening modalities for early detection. However, existing techniques such as mammography and MRI exhibit limitations regarding sensitivity and specificity. Microwave imaging has recently emerged as a promising technology for breast cancer diagnosis, exploiting the dielectric contrast between normal and malignant tissues. Objectives: This study proposes a novel computational framework integrating thresholding, edge segmentation, and logistic regression to enhance microwave image -based breast tumor delineation. Methodology: The employed algorithm selects optimal features using logistic regression to mitigate the class imbalance between tumor and healthy tissues. Localized density thresholds are applied to identify tumor regions, followed by edge segmentation methods to precisely localize the detected lesions. Results: When evaluated on a dataset of microwave breast images, our approach demonstrated high accuracy for detecting and segmenting malignant tissues. Density thresholds ranging from 0.1 to 0.8 showcase the highest accuracy in detecting breast tumors from these images. Conclusions: The results highlight the potential of the proposed segmentation algorithm to improve the reliability of microwave imaging as an adjunct modality for breast cancer screening. This could promote earlier diagnosis and better clinical outcomes. The proposed framework represents a significant advance in developing robust image processing techniques tailored to emerging medical imaging modalities challenged by class imbalance and low intrinsic contrast.