Investigation of Raman Signals from Healthy and Diabetic Animals

Abstract

Abstract:
Investigation of Raman Signals from Healthy and Diabetic Animals
Diabetes mellitus remains one of the most pressing problems in global healthcare, driving demand for non-invasive, reagent-free diagnostic tools. Raman spectroscopy is a promising candidate, as it probes the molecular vibrational fingerprint of biofluids without labeling and is insensitive to water. This work investigated whether Raman spectra of blood plasma reveal systematic biochemical differences between healthy and diabetic animals, and whether these differences allow automatic, unsupervised classification of disease status.
Blood plasma spectra were collected from five animal groups (Control-1, Control-2, non-diabetic hyperglycemia, Diabetes type 1, and Diabetes type 2) using a Renishaw Raman spectrometer with 785 nm excitation. A multi-stage preprocessing pipeline; polynomial baseline correction, smoothing, and normalization was applied to a dataset of hundreds of spectra. Principal Component Analysis was used for dimensionality reduction, followed by four unsupervised clustering algorithms (K-Means, DBSCAN, Agglomerative Clustering, Gaussian Mixture Model), with K-Means giving the best separation.
Clustering recovered the true disease label with high agreement (Adjusted Rand Index > 0.4, Normalized Mutual Information > 0.3). Analysis of PCA loadings and difference spectra identified key marker wavenumbers at 1000 - 1150, 1445, and 1650 cm⁻¹, corresponding to glucose, lipid, and protein bands respectively; differences at these markers were statistically significant (Welch's t-test, p < 0.05).
These results show that Raman spectroscopy of blood plasma captures reproducible, diabetes-related biochemical signatures detectable by unsupervised methods, supporting its use as a foundation for non-invasive diagnostic tools.

Speaker

Asiedu Da-Costa Aboagye
Saratov State University, Saratov, Russia
Ghana

Discussion

Ask question