Linked independent component analysis for multimodal data fusion

FMRIB (Oxford University Centre for Functional MRI of the Brain), Department Clinical Neurology, University of Oxford, Oxford, UK.
NeuroImage (Impact Factor: 6.13). 10/2010; 54(3):2198-217. DOI: 10.1016/j.neuroimage.2010.09.073
Source: PubMed

ABSTRACT In recent years, neuroimaging studies have increasingly been acquiring multiple modalities of data and searching for task- or disease-related changes in each modality separately. A major challenge in analysis is to find systematic approaches for fusing these differing data types together to automatically find patterns of related changes across multiple modalities, when they exist. Independent Component Analysis (ICA) is a popular unsupervised learning method that can be used to find the modes of variation in neuroimaging data across a group of subjects. When multimodal data is acquired for the subjects, ICA is typically performed separately on each modality, leading to incompatible decompositions across modalities. Using a modular Bayesian framework, we develop a novel "Linked ICA" model for simultaneously modelling and discovering common features across multiple modalities, which can potentially have completely different units, signal- and contrast-to-noise ratios, voxel counts, spatial smoothnesses and intensity distributions. Furthermore, this general model can be configured to allow tensor ICA or spatially-concatenated ICA decompositions, or a combination of both at the same time. Linked ICA automatically determines the optimal weighting of each modality, and also can detect single-modality structured components when present. This is a fully probabilistic approach, implemented using Variational Bayes. We evaluate the method on simulated multimodal data sets, as well as on a real data set of Alzheimer's patients and age-matched controls that combines two very different types of structural MRI data: morphological data (grey matter density) and diffusion data (fractional anisotropy, mean diffusivity, and tensor mode).

1 Follower
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The normal myelination of neuronal axons is essential to neurodevelopment, allowing fast inter-neuronal communication. The most dynamic period of myelination occurs in the first few years of life, in concert with a dramatic increase in cognitive abilities. How these processes relate, however, is still unclear. Here we aimed to use a data-driven technique to parcellate developing white matter into regions with consistent white matter growth trajectories and investigate how these regions related to cognitive development. In a large sample of 183 children aged 3 months to 4 years, we calculated whole brain myelin volume fraction (VFM ) maps using quantitative multicomponent relaxometry. We used spatial independent component analysis (ICA) to blindly segment these quantitative VFM images into anatomically meaningful parcels with distinct developmental trajectories. We further investigated the relationship of these trajectories with standardized cognitive scores in the same children. The resulting components represented a mix of unilateral and bilateral white matter regions (e.g., cortico-spinal tract, genu and splenium of the corpus callosum, white matter underlying the inferior frontal gyrus) as well as structured noise (misregistration, image artifact). The trajectories of these regions were associated with individual differences in cognitive abilities. Specifically, components in white matter underlying frontal and temporal cortices showed significant relationships to expressive and receptive language abilities. Many of these relationships had a significant interaction with age, with VFM becoming more strongly associated with language skills with age. These data provide evidence for a changing coupling between developing myelin and cognitive development. Hum Brain Mapp, 2014. © 2014 Wiley Periodicals, Inc.
    Human Brain Mapping 09/2014; 35(9). DOI:10.1002/hbm.22488 · 6.92 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In the biomedical field, current technology allows for the collection of multiple data modalities from the same subject. In consequence, there is an increasing interest for methods to analyze multi-modal data sets. Methods based on independent component analysis have proven to be effective in jointly analyzing multiple modalities, including brain imaging and genetic data. This paper describes a new algorithm, three-way parallel independent component analysis (3pICA), for jointly identifying genomic loci associated with brain function and structure. The proposed algorithm relies on the use of multi-objective optimization methods to identify correlations among the modalities and maximally independent sources within modality. We test the robustness of the proposed approach by varying the effect size, cross-modality correlation, noise level, and dimensionality of the data. Simulation results suggest that 3p-ICA is robust to data with SNR levels from 0 to 10 dB and effect-sizes from 0 to 3, while presenting its best performance with high cross-modality correlations, and more than one subject per 1,000 variables. In an experimental study with 112 human subjects, the method identified links between a genetic component (pointing to brain function and mental disorder associated genes, including PPP3CC, KCNQ5, and CYP7B1), a functional component related to signal decreases in the default mode network during the task, and a brain structure component indicating increases of gray matter in brain regions of the default mode region. Although such findings need further replication, the simulation and in-vivo results validate the three-way parallel ICA algorithm presented here as a useful tool in biomedical data decomposition applications.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Multi-modal data analysis techniques, such as the Parallel Independent Component Analysis (pICA), are essential in neuroscience, medical imaging and genetic studies. The pICA algorithm allows the simultaneous decomposition of up to two data modalities achieving better performance than separate ICA decompositions and enabling the discovery of links between modalities. However, advances in data acquisition techniques facilitate the collection of more than two data modalities from each subject. Examples of commonly measured modalities include genetic information, structural magnetic resonance imaging (MRI) and functional MRI. In order to take full advantage of the available data, this work extends the pICA approach to incorporate three modalities in one comprehensive analysis. Simulations demonstrate the three-way pICA performance in identifying pairwise links between modalities and estimating independent components which more closely resemble the true sources than components found by pICA or separate ICA analyses. In addition, the three-way pICA algorithm is applied to real experimental data obtained from a study that investigate genetic effects on alcohol dependence. Considered data modalities include functional MRI (contrast images during alcohol exposure paradigm), gray matter concentration images from structural MRI and genetic single nucleotide polymorphism (SNP). The three-way pICA approach identified links between a SNP component (pointing to brain function and mental disorder associated genes, including BDNF, GRIN2B and NRG1), a functional component related to increased activation in the precuneus area, and a gray matter component comprising part of the default mode network and the caudate. Although such findings need further verification, the simulation and in-vivo results validate the three-way pICA algorithm presented here as a useful tool in biomedical data fusion applications.
    NeuroImage 04/2014; 98. DOI:10.1016/j.neuroimage.2014.04.060 · 6.13 Impact Factor


Available from