Multi-Input CNN for molecular classification in breast cancer
Mohamed Gasmi, Makhlouf Derdour, Abdellatif Gahmousse, Mohamed Amroune, Hakim Bendjenna, Brahim Sahraoui
Abstract
Molecular classification in pathological anatomy is an important task as it is extremely convenient for the diagnosis of cancer and its subtypes for adequate therapeutic choice. With the development of computer vision, cancer classification has become an interdisciplinary subject in both medicine and computer vision.A multi-input convolutional neural network is designed for the molecular classification of cancer based on a collected dataset, which contains four tissues treated with four antibodies; each one of them is composed of 33 images. The proposed model achieves a satisfactory accuracy of 90.43% after data augmentation. Even though the data augmentation contributes to the model, the accuracy is still limited by the lack of sample diversity.