Artificial intelligence-enhanced diagnostics for pediatric asthma and respiratory irregularities using deep learning and wearable sensors

Background Pediatric asthma and respiratory irregularities are challenging to diagnose due to symptom variability and environmental influences. Timely and accurate detection is essential to ensure effective management and minimize long-term complications. Traditional diagnostic approaches often fall short in providing continuous, real-time monitoring, creating a need for innovative, technology-driven solutions in pediatric respiratory care. Methods A unique artificial intelligence (AI)-powered diagnostic system for pediatric asthma and respiratory irregularities (AI-PARI) was introduced in this study. The deep learning (DL) methods and wearable senor data were integrated in this AI-PARI. Thus, continual, and real-time monitoring of respiratory conditions are facilitated by this integration in the AI-PARI. Current methods are not effective, as they mostly utilize conventional medical techniques. Crucial diagnostic method was offered by the AI-PARI, as it utilizes lung sound analysis. An effective diagnostic tool named lung sound analysis is utilized by AI-PARI. The dataset from Kaggle named Pulmonary Sound Dataset is utilized. Diverse, labeled sound recordings from various respiratory conditions are all included in this dataset. Then, respiratory rate, oxygen saturation, airflow patterns, and heart rate variability are monitored in a non-invasive way. This dataset is utilized for training these neural networks. Respiratory patterns, medical treatment, and background data like air quality and humidity are all included in this dataset. Real-time insights and predictive analytics are offered by the healthcare tool named cloud-hosted diagnostic dashboard (C-DB); doctors are greatly aided from this C-DB tool. Then, the data of the patient are secured by federated learning (FL). This FL facilitates customized model training over devices. Results In the detection of asthma cases and respiratory abnormalities, the efficiencies of AI models and standard methods like Artificial Intelligence for Intellectual and Developmental Disabilities (AI-IDD), Supervised Machine Learning (SML), Artificial Intelligence Model (AIM) and Artificial Intelligence–Powered Screening Framework (AI-PSF) are evaluated. From the outcomes, it is clear that the suggested AI model attains 95% accuracy in detecting such asthma cases and respiratory abnormalities. Thus, it is clear that AI framework attains superior performance compared to the baseline approaches. Early detection is facilitated by this system, as it offers real-time insights. So, it reduces the need of visiting hospital for an emergency situation. It also offers a scalable, cost-effective solution for improving pediatric asthma management through intelligent, wearable technology. Future directions include enhancing model generalization, integrating multi-sensor data fusion, and expanding real-world clinical validation.