Deep Learning Techniques based Non-invasive detection of fasting Blood Glucose Level measurement using Electrochemical Saliva
European Journal of Molecular & Clinical Medicine,
2020, Volume 7, Issue 3, Pages 1598-1607
AbstractIn Deep learning methods such as automatic encoder, long-term short-term memory (LSTM) and repetitive neural network (RNN), in mixed group of population, fasting blood glucose level (FBGL) was used to detect the BG level. The Indian population is healthy and sick. The appearance of high FBGL from the electrochemical parameters of human saliva, redox potential, pH, concentration of sodium, and calcium ions was evaluated. Samples were taken from 175 randomly particular persons, half of healthy patients and half of those with diabetes. Models were trained with 70% of all data and tested in the remaining set. In every algorithm, the data points were randomly crossed three times before the model was implemented. The effectiveness of machine learning techniques is presented in terms of the four parameters that are statistically significant, the accuracy, the sensitivity, and the F1 score. The proposed analysis shows that the RNN-based deep learning method yields better results. This deep learning technique to measure blood glucose level non-invasively using electrochemical saliva will helps the society to control the diabetes effectively.
- Article View: 80
- PDF Download: 150