Comparative Analysis of AI Models for Early and Accurate Skin Cancer Diagnosis

Abstract

Skin cancer, especially melanoma, is becoming a growing health problem across the world. For therapy to work and for patients to live longer, it is very important to get a diagnosis early and correctly. As AI has become better, many models have been suggested to automatically sort skin lesions from dermoscopic pictures. This article compares classical machine learning (ML), deep learning (DL), and ensemble-based AI models for early detection of skin cancer. When tested on benchmark datasets like ISIC, PH2, and HAM10000, DL models like ResNet50 had 94.3% accuracy and an AUC-ROC of 0.96. The hybrid CNN + XGBoost model did the best, with 95.6 % accuracy and 0.97 AUC. On the other hand, standard models like Support Vector Machines (SVM) and Logistic Regression only got 87.2 % and 85.1 % accuracy, respectively. The findings show that AI models based on ensemble and DL are far better than traditional methods. They provide more accurate and reliable diagnoses for use in clinical settings.

Ähnliche Arbeiten