Preeti
Principal component analysis (PCA) was used to analyze Raman spectral data, retaining 85% of the explained total variance in the principal components. The explained variance ratio attribute was utilized to determine the variance explained by each principal component. Spectral data was standardized using z-score and combined with PCA for visualization, providing a clearer depiction of data distribution. Figures were included to illustrate the projection direction of data in the principal component space and the distribution of samples along the principal component directions. Additionally, a graph displaying the standard deviation across spectral regions was plotted to visualize the variance of the entire dataset.
Raman spectroscopy of plasma from patients with celiac disease revealed characteristic peaks representing substances like lipids, proteins, nucleic acids, and amino acids. Changes in Raman peaks of proteins and nucleic acids were observed in diseased individuals, along with differences in high-sensitivity C-reactive protein levels and lipoprotein spectra. Various metabolites were found to be lower in CD patients’ serum, indicating differences in lipid composition. The Raman peaks at specific wavelengths reflected abnormalities in lipid metabolism and changes in protein structure or composition in celiac disease patients.
Model evaluation included the assessment of Convolutional Neural Network (CNN), Multi-scale CNN (MCNN), Deep Residual Network (ResNet), and Deep Residual Shrinkage Network (DRSN). DRSN outperformed the other models in accuracy, specificity, sensitivity, and precision, showcasing its enhanced generalization capability in handling spectral data. The evaluation metrics for the test sets of the models were presented, highlighting the superior performance of DRSN in classifying Raman spectral data.
Source link
Source link: https://www.nature.com/articles/s41598-024-64621-4
GIPHY App Key not set. Please check settings