123
In the event that multiple species with multiple m/z values may be involved in a classification, a multivariate analysis can be applied.
Classification is often a goal with IMS data analysis, whether between regions of
tissue, individual substructures within a tissue, or among tissue samples belonging
to different experimental or biological groups. Although unsupervised approaches
can offer some differentiation, supervised approaches can also be used, provided the
data are “labeled”, or the classes are known. The ultimate goal of any supervised
approach is normally to classify new data into a given class using a classification
model built using existing data. Machine learning algorithms such as support vector
machines, neural networks, and random forest can be applied for dimensionality
reduction and classification.
7.12 Common Software for IMS Data Analysis
Cardinal [87] and MSiReader [88] are commonly used open-source R package
and MATLAB software, respectively, for analysis of imaging mass spectrometry
datasets. Cardinal supports MALDI and Desorption Electrospray Ionization
(DESI) IMS workflows. IMS data can be loaded into such software in imzML
format, pre- processed, and analyzed using spatial segmentation, image classification, and statistical analysis functions. As they are open-source, they also support
the development of new computational workflows. Users can create their own
multivariate statistical modeling or model-based visualization using existing
Cardinal functions. Commonly used commercially available software includes
SCiLS and Data Analysis from Bruker, and ImageQuest from ThermoFisher
Scientific.
7.13 Data Registration and Multimodal Approaches
IMS data alone can provide spatially resolved molecular information. However,
coupled with other types of imaging modalities such as microscopy, it allows for
even higher spatial and molecular resolution. For example, IMS provides highly
resolved chemical specificity, but relatively low spatial resolution. Microscopy can
provide a much higher spatial resolution, but relatively low chemical specificity. As
such, combining an IMS dataset with a microscopy image, each obtained on serial
sections of tissue can offer much higher resolution chemical and spatial resolution.
Recent advances to couple these imaging modalities include multimodal image registration [89].
A combination of autofluorescence images, H&E stained images, and IMS
images, pre and post-acquisition can be registered using explicit IMS pixel to laser
ablation marks registration. This method (shown in Fig. 7.8 below) involves a series
7 Matrix-Assisted Laser Desorption/Ionization Imaging Mass Spectrometry…
In the event that multiple species with multiple m/z values may be involved in a classification, a multivariate analysis can be applied.
Classification is often a goal with IMS data analysis, whether between regions of
tissue, individual substructures within a tissue, or among tissue samples belonging
to different experimental or biological groups. Although unsupervised approaches
can offer some differentiation, supervised approaches can also be used, provided the
data are “labeled”, or the classes are known. The ultimate goal of any supervised
approach is normally to classify new data into a given class using a classification
model built using existing data. Machine learning algorithms such as support vector
machines, neural networks, and random forest can be applied for dimensionality
reduction and classification.
7.12 Common Software for IMS Data Analysis
Cardinal [87] and MSiReader [88] are commonly used open-source R package
and MATLAB software, respectively, for analysis of imaging mass spectrometry
datasets. Cardinal supports MALDI and Desorption Electrospray Ionization
(DESI) IMS workflows. IMS data can be loaded into such software in imzML
format, pre- processed, and analyzed using spatial segmentation, image classification, and statistical analysis functions. As they are open-source, they also support
the development of new computational workflows. Users can create their own
multivariate statistical modeling or model-based visualization using existing
Cardinal functions. Commonly used commercially available software includes
SCiLS and Data Analysis from Bruker, and ImageQuest from ThermoFisher
Scientific.
7.13 Data Registration and Multimodal Approaches
IMS data alone can provide spatially resolved molecular information. However,
coupled with other types of imaging modalities such as microscopy, it allows for
even higher spatial and molecular resolution. For example, IMS provides highly
resolved chemical specificity, but relatively low spatial resolution. Microscopy can
provide a much higher spatial resolution, but relatively low chemical specificity. As
such, combining an IMS dataset with a microscopy image, each obtained on serial
sections of tissue can offer much higher resolution chemical and spatial resolution.
Recent advances to couple these imaging modalities include multimodal image registration [89].
A combination of autofluorescence images, H&E stained images, and IMS
images, pre and post-acquisition can be registered using explicit IMS pixel to laser
ablation marks registration. This method (shown in Fig. 7.8 below) involves a series
7 Matrix-Assisted Laser Desorption/Ionization Imaging Mass Spectrometry…