Examining changes in image segmented arteries in response to blood occlusion

The world's expanding and aging population has created a demand for unobtrusive, automated healthcare solutions. Thermal imaging aids in the development of these solutions by enabling the extraction of physiological signals in an unobtrusive manner. This paper aims to examine the potential for thermal video in conjunction with image segmentation methods to quantitatively describe the behavior of blood vessels as blood flow is occluded and relieved. Thermal video of a subject's arm wearing a wireless blood pressure cuff was captured as the cuff operated. Each frame of the thermal video was subjected to three methods of image segmentation (to identify the arteries); basic thresholding, watershed method and level set method. Results from each method were compared; all methods resulted in successful segmentation, but varied in region identification performance. The basic method resulted in the least accurate segmentations, the level set method resulted in the smoothest boundaries and most accurate shape, and the watershed method resulted in a crude shape but best identified the correct areas. When the segmentation results were quantified in terms of region area and examined over time, all methods indicated a smaller area when blood had been occluded, and a larger area both initially and during vascular recovery. The watershed method best characterized the vascular behaviour, while being computationally inexpensive. These results suggest that thermal video in conjunction with image segmentation methods, particularly watershed method, can be used to indicate blood flow through a segmented vessel.