Breast Cancer Detection Using Convolutional Neural Network with EfficientNet Architecture

Breast cancer is the second most deadly disease in Indonesia after lung cancer. This is caused due to a lack of public awareness of this disease. Histopathological examination is one of the methods of detecting breast cancer which is performed manually by a pathologist. However, the diagnostic process of histopathological examination takes a long time and allows a shift in diagnosis. Therefore, technologies are needed to help pathologists to get a diagnosis quickly and accurately. One of them is deep learning. In this research, the deep learning-based system model was developed using the Convolutional Neural Network (CNN) with EfficientNet architecture to help radiologists to get better accuracy and performance in detecting breast cancer. The system performance was tested on 1361 histopathology images consisting of two classes, which are normal and cancer with data ratios for train, validation, and test are 75%, 10%, and 15%. By using the proposed method, the system is able to classify images with the best accuracy rate of 94.26%, loss of 39.52%, precision of 94%, and recall of 94%. This result was achieved on a system model with number parameters, which are image size 64x64 pixels, Adam optimizer, learning rate 0.001, batch size 32, and epoch 50.