Recognition of Urine Smell Patterns in Diabetes Mellitus Patients Using an Artificial Neural Network Based on Semiconductor Sensors and Biosensors (Surface Acoustic Wave)

Diabetes mellitus is a metabolic disease characterized by increased blood sugar levels above normal limits. This disease has a very high risk of death so it needs to be recognized early. Currently, testing blood sugar levels still uses invasive techniques, namely taking examination samples by drawing blood using a syringe, which causes pain in the sufferer's body. This causes some patients to be reluctant to check their blood sugar. Therefore, it is essential to create a tool that can measure blood sugar levels accurately without having to injure the body of diabetes mellitus sufferers, which is called a non-invasive method. Based on these problems and previous research references, this research will develop a tool for recognizing urine odor patterns in Diabetes Mellitus (DM) sufferers by using e-nose equipment. In designing this tool, gas sensors have been used, namely MQ-135, MQ-4, TGS2 2600, TGS 2602, and a Surface Acoustic Wave (SAW) sensor to detect urine smell through the evaporation process. It uses the backpropagation algorithm as an Artificial Neural Network (ANN) method for the process of recognizing patterns produced from steam samples. From the 35 test data carried out, the system could recognize patterns from the urine samples tested with a success rate of 94%. The significance of this research is that it is a tool for early detection of non-invasive DM disease.