Integrated AI and Sensing System for Pulmonary Disease Evaluation
University of Pittsburgh researchers have developed system techniques integrating artificial intelligence (AI) and smartphone sensing, namely PTEase, that can be used as a pulmonary telemedicine device by patients to accurately evaluate pulmonary disease conditions and provide clinically relevant information out of clinic. Using only a commercial smartphone and a 3D-printed attachment with this novel multi-task machine learning (ML) model, it is possible to assess airway measurements from PTEase, extract appropriate biomarkers and determine key clinical information including the probability of disease and lung function indices.
Description
The COVID-19 pandemic highlighted the benefits of telemedicine. However, the risk for patients attending healthcare settings during a respiratory pandemic also demonstrated the clinical need for remote monitoring and evaluation of pulmonary disease. PTEase has been designed to allow for remote monitoring of lung function through measurements of the cross-sectional area (CSA) of each airway segment. Using machine learning (ML), CSA data can be used to determine a vast array of clinical information allowing for accurate remote monitoring of patients with pulmonary diseases.Applications
• Cystic fibrosis (CF)• Asthma
• Pulmonary diseases
Advantages
Current approaches to diagnosis and monitoring pulmonary disease often involve subjective reporting from patients who may fail to recognize early symptoms or slow decline in lung function, leading to acute emergencies. There are existing pulmonary function tests, but these require specialist equipment, are costly (e.g., spirometry can cost >$50,000) and are aerosol-generating, so not suitable during a highly infectious respiratory pandemic.The development of PTEase and the associated mouthpiece uses acoustic signals transmitted from the smartphone and the associated detected reflections along with AI to accurately determine the CSA of a patient’s airway segments. Using ML, it is possible to train the algorithm to make accurate determination of clinical features including details of lung function for clinicians to remotely monitor patients.