Natural Language Processing Approach Using a Neural Network Ensemble (CNN-HSNN) for Skin Cancer and Multi-Disease Classification

Abstract

Dermatological diseases, including skin cancer, represent a significant challenge for global health systems. Early and accurate diagnosis is crucial to improve patient outcomes and reduce treatment costs. This study leverages an ensemble system combining Convolutional Neural Network (CNN) and Hybrid Sequential Neural Network (HSNN) models to accurately classify various dermatological diseases, including skin cancer, Dermatitis Atopica, Melasma (Cloasma), and Vitiligo. The CNN model processes skin cancer data, while the HSNN model handles the other diseases using a combination of embedding, LSTM, and dense layers. The ensemble system achieved a global F1-score of 95.45%, demonstrating balanced diagnostic precision across all diseases. Precision, recall, and F1-scores were consistently high across the different diseases, underscoring the ensemble system's robustness. These results provide a reliable decision-support tool for early diagnosis and personalized treatment of dermatological diseases, ultimately contributing to improved patient outcomes and optimized healthcare efficiency. Future work aims to expand the framework to cover additional dermatological conditions and integrate both text and image data for comprehensive diagnostic analysis.