| Summary: | The transformative impact of Digital Breast Tomosynthesis (DBT) in breast imaging, highlights its three-dimensional reconstruction capabilities to address tissue overlap issues. The emphasis is on the critical importance of accurate microcalcification detection in DBT images for early breast cancer diagnosis. This study investigates the application of deep learning, specifically Region-based Convolutional Neural Networks (RCNN) to enhance microcalcification detection in Digital Breast Tomosynthesis (DBT) images. The study proposes leveraging deep learning in conjunction with DBT to enhance microcalcification detection, capitalising on DBT's unique capabilities. The study aims to address existing gaps in microcalcification detection within DBT and evaluates the potential advantages of a DBT CAD system. The model's performance is thoroughly evaluated across diverse image scenarios, encompassing both noisy (blur) and clean (non-blur) datasets. A comprehensive comparative analysis is conducted, assessing overall performance, detection confidence, and training progress between the two datasets. Results and outcomes showcase the remarkable flexibility and improved microcalcification detection confidence of the RCNN model, even in the presence of image noise. The comparative analysis provides valuable insights into the model's performance under different imaging situations, aiding in informed decision-making for diagnostic purposes. The findings emphasize the critical role of image clarity in improving detection performance and offering valuable insights for future developments in the field. © 2024 IEEE.
|
|---|