Depth-based Patient Monitoring in the NICU with Non-Ideal Camera Placement

Depth cameras can improve the performance of patient monitoring systems without the introduction of multiple sensors in the NICU. A method was developed to correct non-ideal camera placement. The mean absolute percentage error of the method tested on 28 patients was 5.58 for camera angles up to 38.58° away from the optimal camera placement. An ROI selection method was developed and tested for the use of extracting a respiratory rate signal. The ROI selection method was found to have an average Sørensen-Dice coefficient of 0.62 and Jaccard index of 0.46. The signal was compared to a simpler method resulting in an improvement to the percentage of acceptable estimates. An intervention detection method was developed using a vision transformer model, and the performance was compared to the state-of-the-art in the field. The best model was found to achieve a sensitivity of 85.6%, precision of 89.8%, and F1-Score of 87.6%.