Evaluation of novel respiratory sensor in neonatal respiratory monitoring

<bold>Introduction:</bold> There is a need to improve respiratory monitoring for neonatal patients, enabling detection of apnoeas and clinical deterioration. New wearable technology is being developed using artificial intelligence to enhance diagnostic accuracy. <bold>Objectives:</bold> Pre-clinical evaluation of a novel, accelerometer-based respiratory sensor (Pneumowave, Pneumowave Ltd, UK) to capture clinically relevant extremes of respiratory parameters in neonatal patients. <bold>Methods:</bold> Pneumowave sensor was placed on a term neonatal mannequin’s chest during a protocol of mechanical ventilation settings. MATLAB allowed accelerometer waveforms to be visualised across a range of breathing frequencies (RR) and tidal volumes (TVs), inspiratory: expiratory ratios, biosensor positions and apnoea simulation. Accelerometer data was compared to ventilator-measured parameters and Pneumowave algorithm predicted RRs. <bold>Results:</bold> Algorithm RR was compared to visual count RR, for ventilator frequencies 3-75 breaths per minute (bpm), R² 0.05549, P 0.3798, Bland-Altman plot bias –15.13, 95% CI [–68.9-36.68]. Removing outliers, ventilator frequencies 20-60bpm demonstrated a strong statistically significant correlation, R² 0.9372, P <0.0001, Bland-Altman plot bias –1.66, 95% CI [–8.56 to 5.45]. RR was accurately identified, +/-2bpm, for ventilator delivered tidal volumes of ≥7.6ml, with accuracy dropping for tidal volume ≤6ml, Bland-Altman plot bias of –9.11, 95% CI [−43.45-25.22]. <bold>Conclusion:</bold> Pneumowave biosensor can accurately measure RR and pattern across a range of neonatal applicable breathing frequencies and TVs. Further research will develop the sensitivity and specificity of algorithms for monitoring respiratory metrics in neonatal patients.