Deep Learning and Optimization Approaches in Convolutional Autoencoder with Dual-Key Transformer Network Based Smart E-Health Application for the Prediction of Tuberculosis Using Serverless Cloud Computing: A Review

Abstract

Tuberculosis (TB) remains one of the most critical global health challenges, necessitating rapid, accurate, and scalable diagnostic systems. Recent advancements in artificial intelligence (AI), particularly deep learning (DL), have revolutionized medical image analysis and disease prediction. This review focuses on the integration of Convolutional Autoencoders (CAE) and dual-key transformer networks within a smart e-health framework deployed over serverless cloud computing environments for TB prediction. CAEs enable efficient feature extraction and dimensionality reduction from high-dimensional medical imaging data, while transformer-based architectures enhance contextual learning and attention mechanisms for improved classification accuracy. The incorporation of dual-key transformer models further strengthens data security and privacy, which are crucial in healthcare applications. Serverless cloud computing facilitates scalable, cost-effective, and real-time processing of large medical datasets without requiring infrastructure management. Recent studies demonstrate that deep learning models, particularly CNN-based architectures, achieve diagnostic accuracies exceeding 90% in TB detection tasks . Furthermore, cloud-based AI systems have shown significant improvements in accessibility and deployment efficiency in remote healthcare settings. This paper presents a comprehensive review of methodologies, optimization strategies, and system architectures, highlighting current challenges and future research directions in AI-driven TB prediction systems.