Fusion-Based Deep Learning Approach for Skin Cancer Detection Using DenseNet-121 and Transfer Learning

Main Article Content

Hamsalekha R., Glan Devadhas George, T Y Satheesha

Abstract

Skin Cancer is major type of cancer existing worldwide. In-order to improve the survival rates of infected person it is necessary to detect it in the Early stage. Earlier Traditional methods were used to detected the cancer but was more Time consuming and involved complex procedures like Biopsy, and to produce the output consumes more time. With the increase in latest technology in deep learning techniques like convolutional neural networks (CNNs) and other methods has increased the accuracy of detection of Melanoma skin cancer detection. This Research paper presents an approach which uses DenseNet-121 for classification of skin lesions into malignant and benign for binary categories. This methodology also employed transfer learning, data augmentation, and optimization techniques which helps in improved classification performance. The Proposed model achieves accuracy of 91%, Precision: 0.89, Recall (Sensitivity): 0.91, F1 Score: 0.90 on a custom dataset, that helps automating procedures that helps for assisting dermatologists in diagnosis and Treatment. This demonstrates a high performance in contrast to the performance of other cutting-edge networks. As a result, the proposed Deep Learning technique provide a less complex and more cutting-edge model pertaining to automatic detection and identification of melanoma skin Cancer, hence this method increases the likelihood of successfully saving a Person life.


The main aim of this research is to enhance binary skin cancer classification by using transfer learning with DenseNet-121, optimizing the hyperparameters, and applying data augmentation techniques to achieve robust and Accurate performance.

Article Details

Section
Articles