Chronic Obstructive Pulmonary Disease (COPD) is a leading cause of morbidity and mortality worldwide, yet early detection remains challenging-especially in primary care settings where spirometry, the diagnostic gold standard, is often unavailable. This study aims to develop and validate a non-invasive, low-cost COPD screening tool based on artificial intelligence (AI) analysis of cough sounds. Using smartphone-recorded cough audio and clinical data from both COPD patients and non-COPD controls, the investigators will train and test an AI model to identify acoustic signatures associated with COPD. The model will be developed using a prospective cohort from Sir Run Run Shaw Hospital, Zhejiang University School of Medicine, and externally validated in a community-based cohort across nine districts/counties in Zhejiang Province, China.
This is a prospective observational study with a "single-center modeling + external validation" design. Two cohorts will be enrolled: (1) individuals diagnosed with COPD according to the GOLD 2024 criteria, and (2) individuals clinically confirmed as non-COPD. All participants must be ≥18 years old and able to perform a voluntary cough. Each participant will undergo standard clinical assessments-including spirometry (FEV₁, FVC, FEV₁/FVC ratio), CT imaging, blood tests, and a structured questionnaire on smoking history, respiratory symptoms, and risk factors-and will provide a 5-second cough recording via a smartphone. Audio data will be de-identified and used by Xunsheng Medical Technology Co., Ltd. to develop an AI-based screening algorithm. The primary performance metrics (sensitivity, specificity) of the cough sound model will be compared against traditional screening questionnaires using spirometry as the reference standard. The study aims to enroll approximately 3,000 participants to achieve \>90% statistical power in detecting a 10% improvement in sensitivity over questionnaire-based screening.
Study Type
OBSERVATIONAL
Enrollment
3,000
Sir Run Run shaw Hospital Zhejiang University
Hangzhou, Zhejiang, China
Diagnostic Accuracy of the AI-Based Cough Sound Model for Detecting COPD
Sensitivity and specificity of the artificial intelligence (AI) model in identifying individuals with chronic obstructive pulmonary disease (COPD), using post-bronchodilator spirometry (FEV₁/FVC \< 0.70 according to GOLD 2024 criteria) as the reference standard.
Time frame: At the time of enrollment (single visit, baseline assessment)
Area Under the Receiver Operating Characteristic Curve (AUC) of the Cough Sound Model
Discriminative performance of the AI model measured by AUC, compared against COPD screening questionnaires
Time frame: Baseline
Positive and Negative Predictive Values (PPV/NPV)
PPV and NPV of the AI cough sound model for COPD detection in both hospital-derived development cohort and community-based external validation cohort.
Time frame: Baseline
Correlation Between Acoustic Features and COPD Severity
Association between extracted cough acoustic biomarkers (e.g., spectral entropy, pitch, duration, harmonic-to-noise ratio) and COPD severity stages (GOLD 1-4), assessed via linear or ordinal regression models.
Time frame: Baseline
Model Performance Across Subgroups
Sensitivity and specificity of the AI model stratified by age (\<65 vs ≥65 years), smoking status (current/former/never), and presence of comorbid respiratory conditions (e.g., asthma, bronchiectasis).
Time frame: Baseline
Feasibility of Smartphone-Based Cough Recording
Proportion of participants able to successfully complete a high-quality 5-second voluntary cough recording using a standard smartphone under real-world primary care or hospital settings.
Time frame: Baseline
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.