| dc.description.abstract | Mass spectrometry imaging (MSI) is becoming a powerful tool in the bioanalytical studies owing to its unique capability to sensitively map the spatial distribution of broad ranges of molecules on biological samples. Due to the large size and complex structure of the image datasets, conventional analysis methods, such as directly mapping the selected ions, is insufficient to achieve comprehensive data analysis. To increase the data analysis efficiency and fully extract the information contained in MS image data, advanced data analysis methods are needed. This dissertation focuses on the studies using the combined Single-probe MSI method, an ambient MSI technique, with advanced data analysis, including multivariate curve resolution (MCR), machine learning (ML), and multi-modal imaging fusion. MCR is a technique to decompose the hyperdimensional dataset into major components, which possess similar spatial distributions, and extract molecules from each component. ML is becoming increasingly popular for analyzing MSI data due to its superior capability to deal with big data. Here, both supervised and unsupervised ML methods were used to segment the MSI data, providing fast and accurate approaches to image segmentation. In addition, image fusion technique was applied to enhance the MSI data analysis, in which microscope images and MS images were fused together to increase the spatial resolution and correlate the spatial information of protein biomarkers and metabolites. The integration of the Single-probe MSI experimental techniques and advanced data analysis methods can potentially benefit fundamental research and broad types of applications such as in drug discovery and studies of disease. | en_US |
