Tags:AlexNet Architecture, Deep Learning, Malignant Melanoma and Visual Geometry Group
Abstract:
The application of machine learning and deep learning has over the years revealed a high level of image accuracy in medical diagnosis and classifications. This study was focused on the application of deep learning in the staging of Malignant Melanoma coupled with the use of VGG-16 and AlexNet architecture in the training over different datasets, for comparison purposes. The first objective of the study was to analyze the impact of deep learning in Melanoma identification, this was accompanied with the acquisition of datasets for analysis. The dataset used in this study consists of 5342 images that are classified as melanoma and non-melanoma, of which 1000 images were used in each classe. Two models (AlexNet and VGG-16) were used for classification which consist of melanoma and non-melanoma as well as 7 data sets consisting of actinic keratosis; Basal cell carcinoma (bcc), benign keratosis-like lesions, dermatofibroma, Melanoma, Melanocytic nevi and Vascular lesions. The result from this study showed a highly precise result with an accuracy of 96.3%, recall (96.5%), specificity (96%), precision (96%), and a F1 score (96.2%). For the 7 dataset model, AlexNet had an accuracy of 90.2%. Experimentally, the result obtained from this study has revealed that the possible use of deep learning in Malignant Melanoma detection is a highly effective and accurate technique.
Medical Application of Deep Learning-Based Detection on Malignant Melanoma