Breathalyzer 2.0: Non-Invasive Diabetes Detection Using Machine Learning

The Diabetic Patient count is drastically increased nowadays, Diabetes is a long term condition in which the person’s body cannot breakdown the blood sugar adequately due to shortage in the production of insulin. Diabetes should be monitored properly to reduce critical emergencies such as Hyperglycemia and Hypoglycemia. Previous method used to detect diabetes by pricking fingers and by testing it with glucometer which is a invasive method for continuous monitoring, frequent pricking will become painful process. To avoid this, non-invasive method is used. Non Invasive Diabetes detection based on the variation of Acetone level from the exhale breath of the patient. Volatile organic compound(VOCs) such as acetone, ethanol, toluene, Benzene, formaldehyde causes variation in blood glucose level and insulin level. Acetone from the exhale breath provides a direct changes in blood glucose and insulin levels. From that the blood glucose variation can be measured directly based on the change in acetone level .The research discuss on non-invasive techniques to measure the blood glucose level from the exhale breath. Machine learning model is trained using support vector regression(SVR) with a small range of dataset to provide a higher accuracy and to reduce error in the outcome.