Multivariate Analysis of Abdominal Aortic Aneurysm in Mice using Principal Component Analysis and Gaussian Mixture Model

Thumbnail Image



Journal Title

Journal ISSN

Volume Title


The Ohio State University

Research Projects

Organizational Units

Journal Issue


Fourier transform infrared (FT-IR) microspectroscopy enables high spatial resolution biochemical analysis and imaging of tissue. This research utilizes principal component analysis (PCA) and Gaussian mixture models (GMM) to analyze murine abdominal aorta samples of both healthy and diseased tissue with an abdominal aortic aneurysm (AAA) present. The samples were mounted onto glass slides and analyzed via reflectance FT-IR microspectroscopy. The spectra collected per sample numbered in the thousands, with hundreds of data points contained within a spectrum. The size of the hyperspectral data sets and their continuous nature require analysis by robust multivariate techniques like PCA. PCA is widely used to reduce multivariate data into a few dimensions (PCs) that incorporate most of the variance in the data. For more effective analysis of the tissue, GMM was used following PCA to separate tissue spectra from spectra of the glass slide. GMM is an unsupervised clustering technique that assumes a finite number of Gaussian distributions within a data set. Prior application of PCA was found to be critical to successful GMM clustering of the data for removal of background data. GMM clustering was subsequently applied to isolated tissue spectra preprocessed with PCA for various infrared regions to determine which bands were useful for distinguishing healthy and diseased samples. PC 2 provided more accurate clustering of tissue spectra into 2 classes of diseased and healthy. The technique advanced by this research is useful for determining infrared bands useful for distinguishing healthy arterial tissue and tissue affected by AAA.



multivariate analysis, abdominal aortic aneurysm, principal component analysis, gaussian mixture model, FT-IR spectroscopy