Estimation of Respiratory Volumes During Tidal Breathing Using Two Depth Cameras

Recent advances in depth camera technology enable precise quantification of shape changes of the human torso during breathing and therefore predict respiratory volumes. The aim of this study was to investigate the potential of estimating tidal volumes during relaxed breathing from two depth cameras. Data were analysed from 21 healthy participants. For each participant, a set of 30 physical markers were placed across the torso and tracked using two depth cameras. Marker positions were mapped to 3D coordinates and used as inputs for a volume algorithm that estimated respiratory volumes. A breath-by-breath analysis was then conducted to compare estimated volumes with spirometry measurements. The camera-derived volumes were strongly correlated with spirometry (marginal R² = 0.941) and showed excellent agreement with a mean (SD) absolute difference of 78.0 (34.8) mL. Bland-Altman analysis showed a bias of -4.6 mL and 95% limits of agreement of -192.3 to 183 mL. The reduced 17-marker configuration also showed a strong correlation with spirometry (marginal R² = 0.918) but reduced agreement, with a mean (SD) absolute difference of 107.4 (58.2) mL and a Bland-Altman bias of -76.4 mL (95% limits of agreement: -298.1 to 145.3 mL). This is the first study to use a marker-based system to estimate respiratory volumes from two facing depth cameras. The use of markers allows for segmentation of the torso and therefore offers the potential to characterise breathing without the need for an expensive laboratory motion capture system.