Abstract: Breast cancer (BC) affects women with high incidence and death rates. Its diagnosis is achieved through medical imaging, requires the analysis of skilled radiologists, and is subject to human errors. This research utilizes Machine Learning (ML) for automatic BC classification with a focus on clinical context and standards for Artificial Intelligence (AI) in medicine. To limit the effort required by humans in analyzing medical images, Convolutional Neural Networks (CNN) are used to automatically extract features from preprocessed regions of interest (ROI). However, since open access medical data are scarce, Transfer Learning (TL) using pre trained VGG16 was selected and modified to classify the images as normal, benign or malignant. The model’s performance is assessed using the appropriate performance metrics for imbalanced datasets. The results are reported based on Balanced Accuracy, Precision, Recall and F1 score. The model resulted in an Accuracy of 86% with a Balanced Accuracy of 73.33%.

Keywords: Breast cancer, multiclass classification, ROI, VGG16, image preprocessing, transfer learning

Pages: 489-496