SERS spectroscopy with machine learning to analyze human plasma derived sEVs for coronary artery disease diagnosis and prognosis

Authors

Xi Huang, University of Nebraska-LincolnFollow
Bo Liu, University of Nebraska Medical Center
Shenghan Guo, The State University of New Jersey, Arizona State University
Weihong Guo, The State University of New Jersey
Ke Liao, University of Nebraska Medical Center
Guoku Hu, University of Nebraska Medical Center
Wen Shi, University of Nebraska Medical Center
Mitchell Kuss, University of Nebraska Medical Center
Michael J. Duryee, University of Nebraska Medical Center
Daniel R. Anderson, University of Nebraska Medical CenterFollow
Yongfeng Lu, University of Nebraska - LincolnFollow
Bin Duan, University of Nebraska Medical Center, University of Nebraska-LincolnFollow

Document Type

Article

Date of this Version

9-18-2022

Citation

Bioeng Transl Med. 2022;e10420. https://doi.org/10.1002/btm2.10420

Comments

Open access

Abstract

Coronary artery disease (CAD) is one of the major cardiovascular diseases and represents the leading causes of global mortality. Developing new diagnostic and therapeutic approaches for CAD treatment are critically needed, especially for an early accurate CAD detection and further timely intervention. In this study, we successfully isolated human plasma small extracellular vesicles (sEVs) from four stages of CAD patients, that is, healthy control, stable plaque, non-ST-elevation myocardial infarction, and ST-elevation myocardial infarction. Surface-enhanced Raman scattering (SERS) measurement in conjunction with five machine learning approaches, including Quadratic Discriminant Analysis, Support Vector Machine (SVM), K-Nearest Neighbor, Artificial Neural network, were then applied for the classification and prediction of the sEV samples. Among these five approaches, the overall accuracy of SVM shows the best predication results on both early CAD detection (86.4%) and overall prediction (92.3%). SVM also possesses the highest sensitivity (97.69%) and specificity (95.7%). Thus, our study demonstrates a promising strategy for noninvasive, safe, and high accurate diagnosis for CAD early detection.