Automation of Chronic Heart Failure Diagnosis Using Raman Spectroscopy and Machine Learning Methods

Abstract

Chronic heart failure (CHF) remains one of the most prevalent cardiovascular diseases worldwide. Timely diagnosis of CHF is crucial for effective therapy and prevention of complications. However, existing diagnostic methods require significant resources and may fail to detect the disease at early stages. Therefore, the development of new, non-invasive, and highly sensitive diagnostic approaches, particularly those based on Raman spectroscopy, is an urgent research direction [1].
In this study, in vitro blood serum spectra were obtained using surface-enhanced Raman spectroscopy (SERS), as described in detail in [2]. A total of 229 samples were analyzed, including 180 from patients with confirmed CHF and 49 from a control group. To enhance the informativeness of the spectra and reduce data dimensionality, the multivariate curve resolution alternating least squares (MCR-ALS) algorithm was applied. This method enabled the effective extraction of biochemically significant spectral components associated with proteins, lipids, and metabolites.
The concentration profiles of the extracted components were then used as features for constructing machine learning models: logistic regression, support vector machine, random forest, and gradient boosting. Model performance was evaluated using 5-fold cross-validation, ROC curves, and accuracy metrics.
The highest classification performance was achieved using an ensemble method combining several models via stacking. In addition to concentration profiles, clinical data such as systolic and diastolic blood pressure, end-systolic dimension, total bilirubin, and hematocrit were used as supplementary features for classification.
The application of modern machine learning methods opens up opportunities for using the model in screening and outpatient settings. The proposed intelligent analysis method for blood spectra based on SERS and MCR-ALS in combination with machine learning represents a promising tool for early CHF diagnosis.

[1] C. G. Atkins, K. Buckley, M. W. Blades, R. F. Turner, Raman spectroscopy of blood and blood components, Applied Spectroscopy, vol. 71, pp. 767-793, (2017).
[2] S. Z. Al-Sammarraie, L. A. Bratchenko, E. N. Typikova, V. P. Zakharov, I. A. Bratchenko, P. A. Lebedev, Silver nanoparticles-based substrate for blood serum analysis under 785 nm laser excitation, Journal of Biomedical Photonics & Engineering, vol. 8, pp. 010301, (2022).

Speaker

Irina Pimenova
Samara University
Russia

Discussion

Ask question