Analysis of glucose level detection device performance by chemometrics approached and diffuse reflectance fourier transform near infrared spectroscopy / Noor Nazurah Mohd Yatim
Diabetes mellitus is a chronic disease attributed by body that experiences abnormal insulin production that causes high level of glucose in blood. There are three types of blood glucose monitoring which are invasive, minimal invasive and non-invasive. Current glucose measurement method is an invasiv...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/38437/1/38437.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Diabetes mellitus is a chronic disease attributed by body that experiences abnormal insulin production that causes high level of glucose in blood. There are three types of blood glucose monitoring which are invasive, minimal invasive and non-invasive. Current glucose measurement method is an invasive type of measurement that require draw out blood sample several times a day with pain and stressful feeling. Meanwhile, non-invasive optical glucose detection is a painless and harmless method that has been shown in this study to help patients monitor their glucose level and improved patient's health. Therefore, the initial investigations were conducted to evaluate the potential of FT-NIRS spectra towards non-invasive level glucose detection using chemometrics analysis. An optical device has been developed for the measurement of glucose level which consists of tungsten halogen light source, Fourier Transform Near Infrared (FTNIR) with photodetector, bifurcated optical fibre as a waveguide, NIR cuvette sample holder and white reflectance standard. Samples in this study were glucose in water, glucose in intralipid, human skin and intracardiac rat's blood. The analysis tools to analyse all these data were Principal Component Analysis (PCA) to analysis and classify types of sample and Partial Least Squared (PLS) to predict glucose concentration. The glucose solutions with different pH, glucose in Allura red and human body skin were classified by PCA. Meanwhile, PLS regression model gave good prediction values for the glucose level in water, glucose in intralipid and rat's intracardiac blood sample. The PLS regression model developed was validated by Root Mean Square Error (RMSE) and Coefficient of Determination (R2 ). The best PLS model was achieved by using Multiplicative Signal Correction (MSC) data preprocessing for the glucose level detection in water which are 125 mg/dl (6.9 mmol/1) RMSE and 0.55 R2 . Meanwhile, glucose in intralipid model by Savitzky Golay (SG) showed 16.3 mg/dl (0.9 mmol/1) RMSE and 0.99 R2 . As for intracardiac rat's blood, the best model was obtained with the application of SG since it provide the lowest RMSE which is 2.05 mg/dl (0.11 mmol/1) and highest value of R2 which is 0.96. This analysis showed promising results for application on human. |
|---|