A hierarchical model for probabilistic independent component analysis of multi-subject fMRI studies.
Ying Guo, Li-Tang
Biometrics August 22, 2013 DOI: 10.1111/biom.12068 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper Peer reviewed |
|---|---|
| Topics | Meditation |
| Keywords | Humans Magnetic resonance imaging Brain mapping Data interpretation, statistical Models, statistical Principal component analysis Models, neurological Computer simulation Maximum likelihood estimation Em algorithm Hierarchical model Independent component analysis Group inferences Multi-subject imaging data |
| Key points | Proposes a hierarchical probabilistic group ICA method that models subject-specific temporal and spatial effects in multi-subject fMRI data, providing population- and subject-level network estimates and a formal statistical framework for comparing networks across subjects. The authors argue existing group ICA methods ignore between-subject spatial variability. An EM algorithm with explicit E- and M-step forms is developed, and the method is compared with two popular group ICA methods via simulations and illustrated on Zen meditation fMRI data. |
Abstract
An important goal in fMRI studies is to decompose the observed series of brain images to identify and characterize underlying brain functional networks. Independent component analysis (ICA) has been shown to be a powerful computational tool for this purpose. Classic ICA has been successfully applied to single-subject fMRI data. The extension of ICA to group inferences in neuroimaging studies, however, is challenging due to the unavailability of a pre-specified group design matrix. Existing group ICA methods generally concatenate observed fMRI data across subjects on the temporal domain and then decompose multi-subject data in a similar manner to single-subject ICA. The major limitation of existing methods is that they ignore between-subject variability in spatial distributions of brain functional networks in group ICA. In this article, we propose a new hierarchical probabilistic group ICA method to formally model subject-specific effects in both temporal and spatial domains when decomposing multi-subject fMRI data. The proposed method provides model-based estimation of brain functional networks at both the population and subject level. An important advantage of the hierarchical model is that it provides a formal statistical framework to investigate similarities and differences in brain functional networks across subjects, for example, subjects with mental disorders or neurodegenerative diseases such as Parkinson's as compared to normal subjects. We develop an EM algorithm for model estimation where both the E-step and M-step have explicit forms. We compare the performance of the proposed hierarchical model with that of two popular group ICA methods via simulation studies. We illustrate our method with application to an fMRI study of Zen meditation.
Comparable studies
Other theoretical and historical papers on meditation, most cited first.
| Study | Year | Design | Participants |
|---|---|---|---|
| Mindfulness: A proposed operational definition. | 2004 | Theoretical or philosophical paper | |
| Mechanisms of mindfulness | 2005 | Theoretical or philosophical paper | |
| How Does Mindfulness Meditation Work? Proposing Mechanisms of Action From a Conceptual and Neural Perspective | 2011 | Theoretical or philosophical paper | |
| Mindfulness-based cognitive therapy for depression | 2006 | Theoretical or philosophical paper | |
| Self-awareness, self-regulation, and self-transcendence (S-ART): a framework for understanding the neurobiological mechanisms of mindfulness | 2012 | Theoretical or philosophical paper |