Noninvasive diabetes mellitus detection using facial block color with a sparse representation classifier

Diabetes mellitus (DM) is gradually becoming an epidemic, affecting almost every single country. This has placed a tremendous amount of burden on governments and healthcare officials. In this paper, we propose a new noninvasive method to detect DM based on facial block color features with a sparse representation classifier (SRC). A noninvasive capture device with image correction is initially used to capture a facial image consisting of four facial blocks strategically placed around the face. Six centroids from a facial color gamut are applied to calculate the facial color features of each block. This means that a given facial block can be represented by its facial color features. For SRC, two subdictionaries, a Healthy facial color features subdictionary and DM facial color features subdictionary, are employed in the SRC process. Experimental results are shown for a dataset consisting of 142 Healthy and 284 DM samples. Using a combination of the facial blocks, the SRC can distinguish Healthy and DM classes with an average accuracy of 97.54%.