- Research Article
- 10.1080/00387010.2024.2392713
Two-dimensional low-field nuclear magnetic resonance approach for the detection of metabolic syndrome in human serum
- Aug 19, 2024
- Spectroscopy Letters
- Yuchen Wu + 4 more +4
Metabolic syndrome (MetS) has become a major public health challenge in recent years. Nuclear magnetic resonance, as a nondestructive measurement, has been widely used in medical diagnosis and screening. Compared to conventional magnetic resonance imaging and nuclear magnetic spectroscopy, low-field nuclear magnetic resonance is superior in portability, simplicity, and low-cost. In this study, a novel approach based on low-field magnetic resonance analysis is proposed to detect metabolic syndrome. T1–T2 correlation relaxometry is employed to measure human serum. Combining with a classification model based on partial least squares discriminant analysis (PLS-DA), the detection of metabolic syndrome is realized. The classification models based on both T1–T2 relaxation correlation data and spectrum are compared. Additionally, classification model based on convolutional neural networks (CNNs) is also evaluated for comparison. The experimental results indicate that the PLS-DA model based on the T1–T2 spectrum achieves a better performance. Specifically, the area under curve, accuracy, F1 score, recall and precision of the classification model evaluated on testing group are 0.784, 70.91%, 75.00%, 77.42%, and 72.73%, respectively. On the other hand, the testing results of CNN model trained using T1–T2 spectrum are 0.774, 65.45%, 61.22%, 48.39%, and 83.33% for AUC, accuracy, F1 score, recall, and precision. The experimental results reveal the limitation of CNN models on small datasets. In this study, T1–T2 correlation relaxometry is employed for the analysis of MetS for the first time, and the feasibility of proposed method has been demonstrated.
Read more