Skin Cancer Detection with Severity Analysis Using Machine and Deep Learning Techniques

View at publisher Summarise with VIRAT
Authors: K. Deenu, Kaviya S, G. Aravindh, D Arokiamercy, G Devadharshini, A Hemambika
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper describes a novel system for systemically detecting skin cancer and stage of severity using a hybrid approach combining machine learning and deep learningbased approaches. The framework uses a modified DenseNet architecture for the classification of skin lesions, a UNet model for accurate segmentation of the lesion, and a Random Forest (RF) algorithm for analyzing severity of the lesion. DenseNet allows for efficient reuse of features and propagation of gradients while UNet provides accurate extraction of the boundary of the affected area. Transfer learning with ImageNet weights improves convergence of training and accuracy of diagnoses. The integration of these models will support early and accurate diagnosis of benign and malignant skin lesions with high levels of interpretation. It was evaluated against the ISIC dermatascopic dataset and overall classification was 98 % for skin lesions, which outperformed CNN-based approaches in terms of specificity, sensitivity, and F1-score. The results of a fully automated diagnostic sysmtem improves performance, and provides detailed information on severity, ultimately allowing dermatologists to make evidencebased decisions.

Fields of Study & Subjects

Cutaneous Melanoma Detection and Management AI in cancer detection Medical Image Segmentation Techniques

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
841–847
Publication date
21 January 2026
Language
EN