An Efficient Hybrid Infrastructure for AI Healthcare System: Skin Cancer Classification

Abstract

This paper proposes an adaptive infrastructure tailored for Artificial Intelligent (AI)-based skin cancer classification, designed to meet the computational demands of both lightweight and heavyweight AI models. Using microservices architecture, Kubernetes orchestration, and event-driven architecture (EDA) with Google Cloud Pub/Sub, the system supports real-time processing across mobile and web platforms. The infrastructure ensures 50 ms average response times, $99.99 \%$ uptime, and a throughput of 1000 requests per second. The integrated AI model achieves $97 \%$ accuracy in distinguishing malignant and benign lesions. The infrastructure’s novelty lies in its seamless integration of scalable AI systems into clinical workflows, addressing critical regulatory and operational challenges in healthcare for early detection and continuous skin cancer monitoring.