Efficient Melanoma Disease Detection by Using Convolutional Neural Network

Most of the skin conditions are brought on by bacterial, viral, allergic, or fungus-related infections. The only way to cure melanoma, one of the more hazardous and rapidly spreading illnesses in the entire world, is with prompt detection. The rapid and precise detection of skin disorders is made possible by the advancement of laser technology and photonics-based medical technologies. The medical equipment required for such a diagnosis is costly and scarce. Therefore, deep learning algorithms aid in the early diagnosis of skin diseases. The categorization of skin disorders relies heavily on feature extraction. Deep learning algorithms have replaced manual tasks like extracting features and data rebuilding for classification, which formerly required human labor. The dataset HAM10000 is made up of pigmented skin lesions labeled with one of seven diagnoses. Melanoma, nevus, and seborrheic keratosis are a few of them. With the use of binary cross-entropy loss and the Adam optimizer (an enhanced form of stochastic gradient descent), the proposed technique preprocessed the images, divided the dataset into training, validation, and testing sets, and trained the convolutional neural network (CNN). Based on the dataset, several skin cancer kinds are described according to sex and age. The model has four convolutional layers with two fully connected layer, and a final output layer with one neuron with sigmoid activation. The proposed approach helps to achieve 92% accuracy in classifying skin diseases.