MACHINE LEARNING-BASED MICROCALCIFICATION DETECTION FROM MAMMOGRAMS: FORMER TECHNIQUES AND CHALLENGES, FUTURE RESEARCH TRENDS
J. Anitha, S. Malathi · Biomedical Engineering Applications Basis and Communications · 2025
Breast cancer is a common disease among women, and its timely detection becomes significant for rapid treatment and prolonged survival rate. Mammography continues as a pivotal screening tool that has contributed to the minimization of breast cancer-related deaths; however, this model has drawbacks that lead to troubles in its efficiency and effectiveness. The microcalcifications appearance is one of the important early symptoms of mammograms, whose broadness varies from 0.1 to 1 mm. The accurate recognition and timely identification of malignant microcalcification can assist in the timely identification, treatment, and diagnosis of breast tumors. However, because of the low contrast and small size relative to the image backgrounds, it is hard and tiresome for specialists to make correct estimations and objectives of the microcalcification. A large range of machine learning techniques has been implemented for the timely recognition of breast cancer. However, only several experiments have utilized deep learning techniques to define the automatic classification of detected lesions on mammography. Hence, this survey concentrates on providing an overview of breast cancer detection with the microcalcification of mammography images for understanding the present challenges and future requirements in the detection of breast cancer. The main objective of reviewing the conventional microcalcification-aided breast cancer detection in mammogram images is to assess the present state-of-the-art breast cancer identification models, recognizing the weaknesses and strengths of distinct techniques. Moreover, this review work highlights the areas for enhancement to build highly efficient, reliable, and accurate diagnostic tools by evaluating their datasets utilized, performance measures, methods utilized, and challenges, highly concentrating on enhancing the timely identification and patient outcomes. Thus, in this work, the techniques supported for the segmentation operations in the mammogram image are discussed for analyzing the necessity of these tasks in breast cancer detection. Moreover, in this research work, the techniques, which have been used in the existing breast cancer detection, are reviewed. Further, various standard performance metrics, such as accuracy, sensitivity, specificity, F1-score, and precision used in the existing works, are also considered in this review work. In addition, the advantages and disadvantages of the existing breast cancer detection works are listed for facilitating future works. In the end, the future gaps and the challenges are offered for modeling the advanced techniques in breast cancer detection with the microcalcification on mammography.