Exhaled Breath Analysis Based Diabetes Detection with k-Nearest Neighbors Classifier.

The diagnosis of early stages of diseases such as lung cancer or diabetes is challenging as those diseases don’t have many noticeable symptoms. The human breath, however, contains many VOCs (volatile organic compounds) that could be used as a clue for conducting the proper test. In this study, a custom breathalyzer device is developed for non-invasive detection and diagnosis of diabetes with the use of a human breath print. An array of 6 MOS sensors is used in the prototype to collect 8 VOCs of the breath print of the subjects. The k-nearest Neighbor algorithm is implemented to determine the likelihood of a disease. In the results, an accuracy of 93.6% was achieved.