This paper presents a study about transfer learning using convolutional neural networks for detecting breast cancer in histopathological images. Transfer learning is a deep learning technique that reuses pre-trained neural network models to perform a particular task, which in this paper is to detect breast cancer in the referred images. Three convolutional architectures were tested: ResNet-18, ResNet-152, and GoogLeNet. The architectures were implemented in Python using PyTorch and trained using GPUs in the Google Colaboratory environment. Moreover, a random image processing stage was used to avoid overfitting, as well. Results indicate that ResNet-152 presents the best results reaching a mean accuracy of 84%, a precision of 90%, and F1 Score of 88%.
CITATION STYLE
e Silva, D. C. S., & Cortes, O. A. C. (2022). On Convolutional Neural Networks and Transfer Learning for Classifying Breast Cancer on Histopathological Images Using GPU. In IFMBE Proceedings (Vol. 83, pp. 1993–1998). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-70601-2_291
Mendeley helps you to discover research relevant for your work.