Desorption electrospray ionization mass spectrometry (DESI-MS) imaging provides valuable information on disease state and grade based on lipid profiles in tissue sections. Multivariate exploratory methods, such as principal component analysis (PCA), provide a comprehensive exploration of the spatial and multivariate chemical information contained in MS hyperspectral images, allowing efficient characterisation of biological tissues [1]. In this study, a multivariate class-modelling approach has been deployed after data compression by principal components to identify neoplastic conditions in human tissues. Cancerous and adjacent normal tissue sections of human bladder tissue were imaged using a commercial LTQ linear ion trap mass spectrometer and a lab-built prototype DESI-MS ion source. Experiments were carried out in the negative ion mode, using a 5 kV spray voltage, a flow rate of 1.5 μL/min, and 150 psi nebulising gas (N2) pressure. The spray solvent was acetonitrile:water (50:50 v/v). Mass spectra were acquired over the m/z range 150-1000. Reduction of the spectral variables (1:4) by a consecutive-window averaging algorithm, normalisation with respect to the total ion current (TIC) and column-centering were performed. Representative training images for normal and tumour tissues were assembled by merging regions of interest (ROIs) for both of the tissue types. Soft independent models of class analogy (SIMCA) and unequal dispersed classes (UNEQ) were applied as class-modelling methods to build models for the normal tissue. Models were applied (after background segmentation) on an external test set of DESI-MS imaged samples, which were previously diagnosed by means of a microscopy examination performed by a pathologist. Results show the suitability of such a supervised strategy to identify human tumour tissue in, which can be used as a valuable tool in cancer diagnosis, complementary to traditional histopathological tissue examination.
A class-modelling approach for detecting neoplasy in human bladder tissue analysed by desi-ms hyperspectral imaging
PIRRO, VALENTINA;
2013-01-01
Abstract
Desorption electrospray ionization mass spectrometry (DESI-MS) imaging provides valuable information on disease state and grade based on lipid profiles in tissue sections. Multivariate exploratory methods, such as principal component analysis (PCA), provide a comprehensive exploration of the spatial and multivariate chemical information contained in MS hyperspectral images, allowing efficient characterisation of biological tissues [1]. In this study, a multivariate class-modelling approach has been deployed after data compression by principal components to identify neoplastic conditions in human tissues. Cancerous and adjacent normal tissue sections of human bladder tissue were imaged using a commercial LTQ linear ion trap mass spectrometer and a lab-built prototype DESI-MS ion source. Experiments were carried out in the negative ion mode, using a 5 kV spray voltage, a flow rate of 1.5 μL/min, and 150 psi nebulising gas (N2) pressure. The spray solvent was acetonitrile:water (50:50 v/v). Mass spectra were acquired over the m/z range 150-1000. Reduction of the spectral variables (1:4) by a consecutive-window averaging algorithm, normalisation with respect to the total ion current (TIC) and column-centering were performed. Representative training images for normal and tumour tissues were assembled by merging regions of interest (ROIs) for both of the tissue types. Soft independent models of class analogy (SIMCA) and unequal dispersed classes (UNEQ) were applied as class-modelling methods to build models for the normal tissue. Models were applied (after background segmentation) on an external test set of DESI-MS imaged samples, which were previously diagnosed by means of a microscopy examination performed by a pathologist. Results show the suitability of such a supervised strategy to identify human tumour tissue in, which can be used as a valuable tool in cancer diagnosis, complementary to traditional histopathological tissue examination.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
