We present the preliminary results from a study investigating the use of thermographic images to identify the peripheral arterial disease (PAD). Our aim is to define the role of thermography as an additional examination in determining the degree of the arterial impairment. We focus on thermograms of the sole. A greater impairment of the leg arteries leads to a reduced blood flow to the sole and, consequently altered surface temperature. We set the initial hypothesis that this change is measurable with a thermogram. We analyse the thermograms by determining the regions of interest on the subject's left and right sole using classical computer vision techniques, including edge detection, contouring, connecting disconnected contours, masking, and manual correction. For each region of interest, we calculate the mean and standard deviation of the surface temperature and compare the discriminative potential of these variables in distinguishing PAD samples from those of the healthy population. For validation, we rely on reference clinical data and morphological examinations of the affected arteries. We analyse 27 thermograms of the soles of patients with PAD. For each of sample, complete occlusion of at least one artery supplying blood to the sole is confirmed by a morphological examination. We also analyse 34 thermograms of the soles of healthy individuals. All thermograms are obtained in a clinical setting. Although statistical differences between the two populations are confirmed for both variables, our preliminary results suggest that the standard deviation has a greater discriminative potential than the mean sole surface temperature. For 83% of the PAD patient samples with arterial occlusions, the standard deviation value is lower than the minimum value observed in the healthy population. The Pearson correlation coefficient between the ankle-brachial pressure index and the mean and standard deviation of the sole surface temperature are 0.20 and 0.63, respectively. Finally, considering the temperature normalisation in this preliminary study does not support a stronger correlation between either the mean or the standard deviation of the surface temperature and the ankle-brachial pressure index. In the future, a much larger study should be conducted to confirm these results. Based on the results of a larger study, a standardised model for the PAD identification could be developed. In addition to defining the expected ranges of standard deviation values of surface temperature for the PAD patients and healthy individuals, temperature distribution patterns in relation to the degree of the arterial impairment could also be investigated. Moreover, it would be extremely valuable to repeat this investigation using thermograms of PAD patients before and after various revascularisation procedures. Finally, with a sufficiently large database of thermograms, future studies should also focus on an automatic recognition of regions of interest using machine and deep learning techniques.
|
