Hybrid CNN-RNN fusion for enhanced breast cancer detection in early stage

One of the most prevalent cancers in women and one with a high death rate is breast cancer. Since determining the actual origin of breast cancer is challenging, early detection of the illness is essential to lowering the death rate from breast cancer-related causes. Recently, the medical industry has benefited from the application of deep learning techniques for disease classification and identification. This research uses an ultrasound image dataset to provide a hybrid deep learning method for breast cancer diagnosis. The model classifies ultrasound images into three classes normal, benign, and malignant. In contrast to the widely used serial technique to extract image features by a convolutional neural network (CNN) and then giving them as input into a recurrent neural network (RNN), our model extracts image features using a structure made up of a CNN and an RNN with LTSM layer and as an extra layer, we have added RELU with softmax. ReLU helps the first hidden layer receive errors from the last layer to adjust all weights between layers and the softmax layer which will divide each class prediction into probabilities and the class with the highest probability will be the best prediction and help in enhancing accuracy. The obtained results indicate that the suggested method outperforms the others in terms of evaluation criteria like an accuracy of 99.35%. In this manner, the proposed hybrid model helps in breast cancer detection.