Abstract Background Traditional spirometry, whilst foundational to respiratory medicine for over a century, presents limitations for continuous monitoring of lung function. The emergence of smartphone-based digital surrogate techniques offers potential solutions for observation of respiratory parameters, particularly crucial for chronic respiratory conditions requiring regular monitoring. Rigorous assessment of parameter agreement (PEF, FEV1, FVC) between conventional clinical devices and smartphone-enhanced video methods is essential. Methods We conducted a prospective observational study involving 85 participants, including healthy volunteers and patients diagnosed with COPD, ILD, and NMD. Participants performed twice-daily measurements of FEV1/FVC ratio, peak expiratory flow (PEF), and forced vital capacity over a defined period. A machine learning approach utilising a four-layer sequential neural network was developed, extracting features from breathing cycles including peak flow gradients and forced expiratory metrics. The model was validated using five-fold cross-validation, with measurements compared against standard spirometry at clinical visits. Results The smartphone-based system demonstrated strong correlation with conventional spirometry values (r = 0.89, p 0.001 for FEV1/FVC) and excellent reproducibility for twice-daily measurements (ICC=0.92). Bland-Altman analysis confirmed agreement within clinically acceptable ranges across all respiratory parameters. The system successfully captured diurnal variations in lung function and effectively identified significant deviations from baseline, demonstrating robust performance suitable for monitoring patients across GOLD 2 to GOLD 3 classifications. In conclusion this study provides evidence that smartphone spirometry achieves clinically acceptable agreement with conventional methods. The simplicity and ease of use support its potential as a valuable tool for home monitoring, early detection of exacerbations, and enabling earlier therapeutic intervention. This abstract is funded by: electronRx
A D Boyle (Fri,) studied this question.