Comparison of Cutting Edge Convolutional Neural Network for Breast Cancer Histopathology Image Diagnosis
Tongyuan Qian · Atlantis Highlights in Computer Sciences/Atlantis highlights in computer sciences · 2023
Female breast cancer, especially the invasive ductal carcinoma, is a very common type of cancer.When breast cancer is diagnosed and treated at an early stage, patients can achieve a higher survival rate.In recent years, neural networks have shown their potential in medical fields.Therefore, this work focused on applying three state-of-the-art convolutional neural networks (CNNs), namely ResNet50V2, InceptionV3 and VGG16 to diagnosing breast cancer from histopathology images to verify whether CNNs can be an effective tool in this case.The three architectures were trained with an original dataset of breast cancer histopathology images.After the image pre-processing and the hyperparameter tuning, the evaluation and comparison of the networks' performance were performed.They were evaluated through several statistical analysis based on accuracy, recall, precision, F1-score and training time.The experimental results showed that InceptionV3 obtained the best performance with the accuracy of 87.12% and F1-score of 86.99.ResNet50V2 achieved a close performance with a 74% training time compared with InceptionV3.The result proved that the state-of-the-art CNNs can be considered as a supportive tool that can help diagnosing breast cancers.