Skin Cancer Classifier: Performance Enhancement Using Deep Learning Models

Skin cancer ranks among the most prevalent human malignancies, primarily diagnosed through visual methods, commencing with clinical screening, followed by dermoscopic analysis. The intricate alterations in the visual appearance of skin lesions pose a considerable challenge for automated classificati...

Full description

Saved in:
Bibliographic Details
Published inInternational Conference on Signal Processing and Communication (Online) pp. 721 - 725
Main Authors Mishra, Swati, Agarwal, Megha
Format Conference Proceeding
LanguageEnglish
Published IEEE 20.02.2025
Subjects
Accuracy
Analytical models
Computational modeling
Convolution neural network
Deep learning
Lesions
Medical Image Analysis
Real-time systems
Residual neural networks
Skin
Skin cancer
Skin lesion cancer
Visualization
Online AccessGet full text

Cover

More Information
Summary:Skin cancer ranks among the most prevalent human malignancies, primarily diagnosed through visual methods, commencing with clinical screening, followed by dermoscopic analysis. The intricate alterations in the visual appearance of skin lesions pose a considerable challenge for automated classification through images. Deep learning techniques have been harnessed to support dermatologists in enhancing diagnostic accuracy. To address this, deep convolutional neural networks (CNNs) are employed to achieve highly discriminating and potentially generalizable results for image classification problems. This research introduces a prediction model designed to classify skin lesions into seven categories using the HAM10000 skin cancer dataset. The MobileNetV2, Resnet50, and DenseNet121 models were used in this study. The MobileNetV2 outperforms with 99.72 % accuracy and achieves 0.88, 0.89, and 0.88 values of precision, recall, and F1 score.
ISSN:2643-444X
DOI:10.1109/ICSC64553.2025.10969043