Automated Breast Cancer Diagnosis Using Deep Learning CNN Models

The study introduces a deep learning model, CCNN, developed to identify and classify eight types of breast cancer: normal adenosis, normal fibroadenoma, normal phyllodes tumor, normal tubular adenoma, malignant ductal carcinoma, malignant lobular carcinoma, malignant mucinous carcinoma, and malignant papillary carcinoma. This model was integrated with five pre-existing models—Xception, InceptionV3, VGG19, MobileNetV2, and ResNet101—all trained for image classification using the ImageNet database, to categorize breast cancer MRI images. The dataset, sourced from Kaggle’s repository, was further augmented using the Generative Adversarial Network (GAN) technique. For 30 experiments carried out to evaluate the introduced model and the transfer learning of five and proposed CCNN models, the F1-score values accuracy yielded were respectively: 98.04%, 96.13%, 98.54%, 97.11%, 94.20% and 98.98%. This study emphasizes the importance of improving the accuracy of early breast cancer detection to enhance the survival rates of patients.