Developing an Automated Breast Cancer Diagnosis Tool from Mammograms Using Deep Learning

The most frequent type of cancer that any person can find in women is breast cancer. To ensure effective treatment and improved health outcomes for patients, early detection of breast cancer is crucial. While mammography is the most common method for screening and detecting breast cancer, it requires skilled radiologists to analyze the images, which can be time-consuming and costly. This can significantly burden individuals in under-resourced countries such as Bangladesh. We intend to create an automated tool for diagnosing breast cancer using deep learning algorithms to tackle this problem. Deep learning (DL), a subset of machine learning, pulls ANNs into action to glean insights from vast datasets. This research seeks to create a web-based tool that accurately and effectively diagnoses breast cancer from mammograms using deep learning. To achieve this, we train the DL system using a publicly available collection containing mammograms. After training the model based on the U-Net architecture with ResNet 18 as an encoder, it can autonomously analyze mammograms and provide a diagnosis by highlighting the regions of interest (ROIs). Our tool achieved a Dice Score of 89%, an IOU score of 80.15% and an F1 score of 88.44 %. Our tool could enrich the accuracy and speed of identifying breast cancer, thereby aiding human radiologists. DLdriven development in the identification of breast cancer has the overall prospect of revamping breast cancer diagnostics and improving patient prognosis in under-resourced countries.