FDLM: An enhanced feature based deep learning model for skin lesion detection

Automated technologies are increasingly widespread in health care and are used to diagnose crucial disorders such as cancer cells. A skin lesion is a kind of skin cancer with benign and malignant components, and early detection is becoming a typical need. Many researchers have established these types of techniques in the past, yet the need for an efficient method still exists to improve the performance of the skin cancer detection process. Deep learning technology is chosen in this research to detect skin lesions from the provided samples. An improved LeNET method is trained with a feature set optimized using the cuckoo search technique. Here, the feature-based deep learning model presents the novelty of the technique designed with various hybrid shape descriptors and compares their performance. Based on accuracy, a feature-based Convolution Neural Network (CNN) with hybrid SURF and ORB has the highest accuracy of 99.62% for skin lesion detection compared to other distinct combinations used in this work. The findings illustrate the usefulness of several hybrid features and their performance with a deep learning model for skin lesion detection.