Deep Learning Approach for Breast Cancer Detection from Histopathology Images

This research introduces a model based on Transfer Learning for the identification of breast cancer via histopathology pictures. The proposed model utilizes the EfficientNetB0 architecture, pre-trained on the ImageNet dataset, with its classification head omitted to facilitate fine-tuning for the specific goal of cancer detection. The Breast Histopathology Images dataset obtained from Kaggle was utilized for training and evaluation. Comprehensive studies validated the model's efficacy, with a training accuracy of 94.49% and a test accuracy of 94.46%, with corresponding losses of 0.13 and 0.16. The model demonstrated exceptional performance on the test data, achieving an accuracy of 94.0%. The confusion matrix reveals a true negative rate of 98.0% and a true positive rate of 74.5%, indicating effective identification of both malignant and non- malignant samples. The findings suggest that the suggested methodology can substantially enhance early breast cancer diagnosis utilizing histopathology data. Subsequent efforts will concentrate on enhancing the model and investigating its implementation in clinical environments for immediate cancer diagnosis.