Clinical diagnosis of diabetes using machine learning and surface-enhanced Raman spectroscopy liquid biopsy: an...

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Clinical Diagnostics

Background:

• Diabetes mellitus poses a growing global health challenge.
• Current diagnostic methods like HbA1c and OGTT may lack early-stage sensitivity and specificity.
• There is a critical need for advanced diagnostic tools for timely diabetes detection.

Purpose of the Study:

• To explore label-free surface-enhanced Raman spectroscopy (SERS) as a novel diagnostic method for diabetes.
• To develop and evaluate a machine learning workflow for analyzing complex SERS spectra from clinical samples.
• To assess the accuracy of SERS combined with machine learning for differentiating healthy individuals from those with diabetes.

Main Methods:

• Utilized label-free SERS with gold nanoparticles for biochemical analysis of clinical serum samples.
• Developed a machine learning workflow incorporating synthetic data augmentation to interpret SERS spectra.
• Tested four distinct machine learning models for classification accuracy.

Main Results:

• Achieved a classification accuracy of 96% for the healthy group.
• Achieved a classification accuracy of 94% for the diabetes group.
• Demonstrated the efficacy of integrating SERS and machine learning for diabetes diagnosis.

Conclusions:

• Label-free SERS coupled with machine learning offers an efficient and accurate method for diabetes diagnosis.
• This liquid biopsy approach provides a non-invasive and rapid alternative to conventional tests.
• The integrated technique holds potential for improving global diabetes detection and patient outcomes.