Classification of glucose concentration in diluted urine using the low-resolution Raman spectroscopy and kernel...

Area of Science:

• Analytical Chemistry
• Biomedical Engineering
• Spectroscopy

Background:

• Accurate detection of glucose in urine is crucial for managing conditions like diabetes.
• Existing methods for urine glucose monitoring can be invasive or lack sensitivity for diluted samples.

Purpose of the Study:

• To develop and validate a non-invasive Raman spectroscopy method for detecting minute glucose concentrations in diluted urine.
• To enhance classification accuracy of normal versus abnormal urine samples using advanced data processing techniques.

Main Methods:

• Simulated diluted urine by diluting normal urine tenfold with water and spiking with glucose (up to 8 mg/dL).
• Utilized low-resolution Raman spectroscopy for data acquisition.
• Applied an optimizing kernel method for data preprocessing and a neural network algorithm for classification.

Main Results:

• The optimizing kernel method significantly improved classification accuracy, increasing it by 92%.
• The combined Raman spectroscopy and neural network approach demonstrated high effectiveness in distinguishing between normal and abnormal urine samples based on glucose levels.

Conclusions:

• Raman spectroscopy, coupled with optimizing kernel and neural network methods, offers a highly accurate and non-invasive approach for detecting glucose in diluted urine.
• This technology holds potential for routine, at-home patient monitoring of urine components, improving healthcare management.