Applying Raman Spectroscopy with Machine Learning to Detect Alzheimer’s Disease

News
Article

A recent study published in the International Journal of Molecular Science shows how Raman spectroscopy combined with machine learning can serve as an improved alternative detection method to preclinical Alzheimer’s diagnosis.

Raman spectroscopy, a popular molecular spectroscopy technique, can be used with machine learning (ML) to better diagnose Alzheimer’s disease, according to a recent study published in the International Journal of Molecular Science (1).

Alzheimer’s disease is a progressive neurodegenerative disorder, and early detection is critical for timely intervention and participation in clinical trials (1,2). Traditional diagnostic methods, which include positron emission tomography (PET), computed tomography (CT), and cerebrospinal fluid (CSF) analysis, often face challenges because of their high costs and invasive nature (1). Another obstacle for using these techniques is that they typically require post-mortem histological examinations for definitive diagnosis (1). As a result, there is a need for diagnostic tools in this field that are more efficient and less invasive.

Alzheimer's disease concept, Elderly woman holding brain symbol of missing jigsaw puzzle, World Alzheimer's, World mental health, Memory loss, Dementia, Parkinson | Image Credit: © ipopba - stock.adobe.com

Alzheimer's disease concept, Elderly woman holding brain symbol of missing jigsaw puzzle, World Alzheimer's, World mental health, Memory loss, Dementia, Parkinson | Image Credit: © ipopba - stock.adobe.com

A research team, led by Andreas Seifert in Spain, explored how to improve diagnostic methods for Alzheimer’s disease. The proposed method put forth in the study utilizes label-free Raman spectroscopy combined with advanced machine learning techniques. The researchers demonstrated in their study that the method could be offering promising alternative to the traditional, more invasive diagnostic methods.

Raman spectroscopy is often used in the clinical space because it provides a molecular fingerprint of a sample's physiology without the need for labeling or complex instrumentation (1). When combined with machine learning algorithms, Raman spectroscopy demonstrated a high potential for distinguishing between healthy individuals and those in the preclinical stage of Alzheimer's disease (1).

One of the highlights of this research study was incorporating patient samples from various cohorts that were sampled and measured over different years (1). This variation added a layer of complexity to the study. The team utilized partial least squares discriminant analysis (PLS-DA) alongside variable selection methods to identify key discriminative molecules, such as nucleic acids, amino acids, proteins, and carbohydrates including taurine/hypotaurine and guanine, from the Raman spectra of dried CSF samples (1).

Despite the inherent sensitivity of Raman spectroscopy to different measurement conditions, the models reliably identified discriminative molecules across different cohorts and years. This consistency underscores the potential of Raman spectroscopy as a reliable tool for preclinical Alzheimer’s diagnosis (1).

The study also demonstrated significant discriminative power, with cross-validated models achieving high discrimination/classification accuracies up to 0.96. The fusion of data from multiple studies not only enhanced the robustness of the overall model but also facilitated a comprehensive assessment of the variables crucial for classification (1). Key wavenumbers identified in the study were consistent with established biomarkers for Alzheimer’s, such as tau proteins and A𝛽42 peptides (1).

As a result, the researchers showed that their model could classify Alzheimer’s disease with great accuracy despite variations in sample cohorts and measurement conditions (1). The implication of their findings is that Raman spectroscopy has a potential place to be used for early Alzheimer's diagnosis.

The researchers also highlighted the next step in their research. They discussed how the goal should be to incorporate a broader range of inter and intraclass variabilities to enhance the model’s robustness and reliability further (1). External validation in clinical settings will be essential to ensure the model's efficacy with new, unseen data (1). Ultimately, the goal is to develop a general, robust prediction model that can be readily adopted in clinical practice, providing a timely and less invasive diagnostic option for patients at risk of Alzheimer's disease.

References

(1) Lopez, E.; Etxebarria-Elezgarai, J.; Garcia-Sebastian, M.; et al. Unlocking Preclinical Alzheimer’s: A Multi-Year Label-Free In Vitro Raman Spectroscopy Study Empowered by Chemometrics. Int. J. Mol. Sci. 2024, 25 (9), 4737. DOI: 10.3390/ijms25094737

(2) Mayo Clinic, Alzheimer’s Disease. Mayo Clinic. Available at: https://www.mayoclinic.org/diseases-conditions/alzheimers-disease/symptoms-causes/syc-20350447 (accessed 2024-06-07).

Related Videos
Related Content