Exploring the Benefits of Data Augmentation for Breast Cancer Classification using Transfer Learning

Breast cancer is the second leading cause of cancer-related fatalities among women. Early detection is crucial for increasing the survival rate of breast cancer, despite its potential for curability. Even after the technological advancements, it is very common that medical practitioners manually inspect these mammograms; this is because it is challenging to implement a Computer Aided Diagnostic (CAD) System due to the unavailability of a large number of quality images from a single source. In this paper, we try to tackle the most common problem in medical mammogram classification, the scarcity of quality images by a unique data augmentation algorithm to improve the dataset, making it unbiased. The technique for the same is also discussed. We also implement a CNN model using Transfer Learning. Deep learning models like DenseNet, ResNet, and EfficientNet have been used with the augmented data. Results from each model have been compared, and conclusions have been drawn by evaluating parameters like Accuracy, Precision, Recall, Sensitivity, Specificity, and F1-Score. Our designed DL model aims to work as a second opinion tool for medical practitioners and hence contribute towards the cause of fighting breast cancer.