Healthcare Cloud Computing-Based Intelligent COVID-19 Detection System with Iot integration using deep learning

Deevi, Durga Praveen and Allur, Naga Sushma and Dondapati, Koteswararao and Chetlapalli, Himabindu and Kodadi, Sharadha and Jayanthi., S (2025) Healthcare Cloud Computing-Based Intelligent COVID-19 Detection System with Iot integration using deep learning. World Journal of Biology Pharmacy and Health Sciences, 21 (3). pp. 433-441. ISSN 2582-5542

The rapidly intertwined technologies of AI, cloud computing, and IoT have brought about a paradigm shift in the healthcare space, with real-time detection of diseases and personalized treatment as one of the major outcomes. Methods available today are either inaccurate or very inefficient when it comes to processing real-time health data streams from IoT devices, not having fault detection mechanisms adequately, being overfitted, and largely incapable of securely managing large volumes of very heterogeneous and unstructured healthcare data. This study presents a cloud-based intelligent COVID-19 detection system that uses DenseNet201-HBO-DNFN to accomplish enhancement on anomaly detection, real-time processing, and predictive analytics in e-healthcare. The system incorporates wearable IoT devices, medical imaging, and deep learning models to process respiratory rates, oxygen saturation, body temperature, and chest X-rays (CXR) into a cloud-based intelligent detection system for COVID-19. Preprocessing techniques, including Histogram Equalization and Min-Max Normalization, ensure improved image quality and standardized input for the model. The DenseNet201 architecture is known for great gradient flow and feature reuse; hence it combines with Hybrid Bayesian Optimization (HBO) and Deep Neuro-Fuzzy Networks (DNFN) to improve diagnostic accuracy. Optimization is done using a parallel computer, distributed file storage, and NoSQL databases in order to perform analysis in real time on very large scales. The new model was trained and validated against the COVID-19 Chest X-ray and CheXpert datasets, achieving a stunning 98.37% accuracy, 98.70% precision, 97.53% recall, and 98.06% F1-score for the end-to-end evolved classifiers to excel. A continuous upward slope in training and validation accuracy graphs was noted to ensure learning effectiveness, whereas loss graphs prove reduced overfitting. Because of this, early disease diagnosis, fault detection, and efficiency prediction in healthcare have been improved - the proposed DenseNet201-HBO-DNFN with much faster interventions and much better real-time analyses improved the patient treatment outcome