Deep Learning Approaches for Microcalcification Detection in Digital Breast Tomosynthesis
K. Jasna, Albert Jerome. S · 2024
Breast cancer is becoming the most prevalent malignancy among Indian women. Mammography, one of the few modern screening methods available, has proven effective for early detection and treatment of nonpalpable, node-negative breast cancers. The advent of digital breast tomosynthesis has changed breast imaging by making it possible to get thin-section, multi-angle X-ray images that are much better at finding microcalcifications and other problems. Recent deep learning techniques have improved detection accuracy, yet challenges persist. Limited labeled datasets, noise interference, and difficulties in managing the high computational demands of 3D data reduce system efficacy. This review examines the deep learning applications used to detect microcalcifications in digital breast tomosynthesis. Microcalcifications are crucial indicators of early-stage breast cancer and are challenging to discern accurately with conventional methods. Deep learning methods and other advanced architectures have shown promise in automatically finding microcalcifications. These methods use the multidimensional data that digital breast tomosynthesis provides. These algorithms excel at learning intricate patterns and spatial relationships within breast tissue, enhancing sensitivity and specificity beyond traditional mammography. The review synthesizes recent advancements, highlighting the performance gains achieved by deep learning models in terms of accuracy, efficiency, and potential clinical impact. The integration of deep learning into digital breast tomosynthesis processes has the capacity to transform breast cancer care by improving early diagnosis and patient outcomes. This can be achieved through learning about model architectures, training methodologies, and conducting comparative evaluations.