|
 |
Noncontact camera-based detection of respiratory motion may support low-burden surveillance of hospitalized dogs, particularly at night. This pilot study evaluated the feasibility of the InsectEye image-analysis application for detecting visually identifiable respiratory motion in selected infrared nighttime video segments. An initial parameter optimization was performed using 13 one-minute video segments from nine dogs. The selected analysis conditions were subsequently evaluated in Experiment 1 using five 5 min video recordings from four dogs. Experiment 2 further assessed respiratory rate estimation using the same validation dataset. InsectEye-detected events were compared with visually confirmed respiratory motion using time-synchronized review. Two criteria were assessed: Both-only, requiring detection of both inspiratory- and expiratory-related motion, and Both+Either, accepting detection of either or both phases. Detection performance improved when the threshold was reduced from 10 to 8, particularly with the 256 × 120 mesh. The 256-120_15_8 condition yielded the highest mean detection rate (87.93 ± 1.87%), fewer false-negative events, and shorter continuous non-detection duration, although false-positive detections increased. The Both+Either criterion performed better than the Both-only criterion. These findings support the feasibility of camera-based respiratory motion detection under selected favorable conditions, but larger studies using physiological respiratory reference methods are required before clinical monitoring or alerting applications can be established.
Research papers (proceedings of international meetings)
