Predicting Malignancy from Breast Histopathological Images Using Deep Neural Networks and Baseline Classifiers
Anindita Mohanta, Sourav Dey Roy, Niharika Nath, Mrinal Kanti Bhowmik · 2024
Cancer is one of the most deadly diseases around the world. Approximately, 38% of the entire population is suffering from cancer. Among various cancers, breast cancer is one of the most prevalent and deadly cancers in women, making it a hot research topic in the field of medicine. The majority of the time, a biopsy approach is used, in which tissue is taken and examined under a microscope. A histopathologist’s lack of training could result in a misdiagnosis. Therefore, computer-aided automatic breast cancer diagnosis systems can help medical experts and pathologists for early diagnosis. From various vision based techniques, Convolutional neural networks (CNNs) have recently emerged as the preferred deep learning techniques for the classification and detection in the medical domain. In this paper, we measured the perception capability of CNNs for classification of histopathological images for breast abnormality detection. We adopted the standard CNNs models as different aspects of transfer learning module, fine-tuning module, and feature extraction module with the support vector machine (SVM) classifier for benign and malignant classification. For effective analysis, separate training and performance assessment is done on the used CNN models at four different image magnifying factors. This study can help to understand the usability of different deep learning approaches in the medical domain.