Automatic detection of periodontitis using intra-oral images

Periodontitis is a chronic inflammatory disease of the supportive tissues and bone surrounding the teeth. In severe cases, it can consequently lead to tooth loss. This disease is most prevalent in rural and remote communities where regular dental visits are limited. Hence, there's a need for a periodontal screening tool for use by allied health professionals outside of dental clinics to detect periodontitis for early referral and intervention. In this paper two algorithms have been proposed and applied on two independently collected datasets in Germany and Australia with 20 and 24 participating subjects respectively; in the first algorithm, intra-oral images of before periodontitis treatment have been considered as diseased subjects and the images of after treatment have been considered as healthy subjects. Using the histogram of pixel intensity as our classification feature, the healthy and diseased subjects have been classified with an accuracy of 66.7%. In the second algorithm, using the difference between the histograms as our classification features, images of “before” and “after” treatment have been classified with an accuracy of 91.6%. If used in a smart phone application, the first algorithm can help people with limited access to dental clinics to be screened for periodontitis by allied health professionals in any healthcare setting. The second algorithm may be useful in helping non-dental personnel to monitor the progress of periodontal treatment.