Extracting ventilatory waveforms from screen recordings: a validated image processing methodology and its application...

Access to high-fidelity ventilatory waveform data remains a significant challenge in respiratory mechanics research, particularly in resource-constrained environments. This study introduces and validates a novel, accessible framework for acquiring ventilatory data by applying image processing techniques to ventilator screen recordings. The accuracy of the framework was evaluated, demonstrating high fidelity (> 0.99) when compared with sensor-derived data from a laboratory emulator. Its reliability was further confirmed by demonstrating that image-derived patient data reproduced established correlations in respiratory mechanics previously obtained from sensor-based measurements. As a proof of concept for the utility of this validated framework, a novel predictive single compartment model was developed to estimate expiratory parameters from inspiratory phase data. By introducing two auxiliary parameters (and), this model established the first predictive relationship for expiratory airway resistance and reference pressure, addressing a key limitation of previous approaches. The predictive model performed consistently in sedated patients but as expected, its linear nature was unable to reproduce the complex dynamics of spontaneous breathing and asynchrony. Overall, this study establishes image processing as a reliable and accessible method for ventilatory data acquisition, warranting further validation across diverse technical conditions and patient cohorts. It demonstrates how this data can support new predictive tools and concludes that while linear models fail to capture asynchrony, this predictive failure itself shows promise as a potential non-invasive indicator for its detection.