SCH: AI-Guided Intraoral Multimodal Active Sound Sensing for Advanced Pulmonary Function Characterization

Respiratory diseases are a leading cause of mortality in the United States, also affecting vulnerable neonates with immature immune systems. Each year, over 300,000 newborns are hospitalized due to respiratory distress syndrome, increasing their risk of chronic lung disease, infections, and cognitive impairments. Current pulmonary monitoring technologies are unsuitable or inadequate for neonates, while respiratory care on infants has contributed to over $7 billion healthcare burden over the past decade. This project introduces a novel, non-invasive active sound sensing approach with artificial intelligence (AI)-based analysis of breathing sounds integrated on a smart pacifier device for real-time, continuous respiratory monitoring, particularly for neonates. We propose a multimodal active sound sensing mechanism that emits sound stimuli and analyzes intraoral breathing sound reflections to assess the cardiopulmonary function. The goals are achieved through the following thrusts (aims). 1) Modeling a new active sound sensing mechanism utilizing breathing sound signal reflections at varying active sound frequencies. 2) Designing a new miniaturized, low-power smart pacifier interface with active/passive sound sensing capabilities for real-time pulmonary assessment. 3) Creating a novel AI-guided analysis pipeline for multimodal active sound sensing of respiratory characteristics using novel deep learning convolutional neural network models (MASS-RespNet) with memory-enabled frequency echo blocks and attention mechanisms, and Internet of Medical Things (IoMT) system integration with edge computing AI capabilities. An open-source breathing sound dataset with active and passive reflections will also be created through the data collected from this study. 4) Conducting feasibility studies in a human subject pilot study with 36 neonates to validate the ideas and approaches. This research advances signal processing, deep learning, AI-driven edge computing, and miniaturized wearable technology, with applications extending beyond respiratory health to neurodegenerative diseases, diabetes, and autonomous systems. The ideas and approach are transformative for neonatal respiratory monitoring, enabling remote, precise spirometry through pacifiers, wearables, and smartphones, extending to broader populations and diseases, reducing hospital visits and healthcare costs. The scalable sound-sensing platform supports longitudinal data collection in non-clinical settings, fostering advancements in smart health monitoring, AI, and biomedical research driving future innovations in smart sensing technology. RELEVANCE (See instructions): This project introduces a new mechanism combining active/passive sensing of breathing sounds with artificial intelligence (AI)-based analysis, integrated on a smart pacifier interface for precise real-time, non-invasive pulmonary function monitoring. Our key innovations include uncovering the biological relevance of intraoral breathing sounds in different sound frequencies to accurately and conveniently determine cardiorespiratory characteristics. This transformative approach enables early detection and advanced diagnostic solutions to improve respiratory health outcomes, particularly for infants.