Classifying Breast Cytological Images using Deep Learning Architectures

Abstract

Breast cancer (BC) is a leading cause of death among women worldwide. It remains a critical challenge, causing over 10 million deaths globally in 2020. Medical images analysis is the most promising research area since it provides facilities for diagnosing several diseases such as breast cancer. The present paper carries out an empirical evaluation of recent deep Convolutional Neural Network (CNN) architectures for a binary classification of breast cytological images based fined tuned versions of seven deep learning techniques: VGG16, VGG19, DenseNet201, InceptionResNetV2, InceptionV3, ResNet50 and MobileNetV2. The empirical evaluations used: (1) four classification performance criteria (accuracy, recall, precision and F1score), (2) Scott Knott (SK) statistical test to select the best cluster of the outperforming architectures, and (3) borda count voting system to rank the best performing architectures. All the evaluations were over the FNAC dataset which contain 212 images. Results showed the potential of deep learning techniques to classify breast cancer in malignant and benign, therefor the findings of this study recommend the use of MobileNetV2 for the classification of the breast cancer cytological images since it gave the best results with an accuracy of