Representation, Pattern Information, and Brain Signatures: From Neurons to Neuroimaging

Abstract

Human neuroimaging research has transitioned from mapping local effects to developing predictive models of mental events that integrate information distributed across multiple brain systems. Here we review work demonstrating how multivariate predictive models have been utilized to provide quantitative, falsifiable predictions; establish mappings between brain and mind with larger effects than traditional approaches; and help explain how the brain represents mental constructs and processes. Although there is increasing progress toward the first two of these goals, models are only beginning to address the latter objective. By explicitly identifying gaps in knowledge, research programs can move deliberately and programmatically toward the goal of identifying brain representations underlying mental states and processes.

Inter-subject pattern analysis: a straightforward and powerful scheme for group-level MVPA

“…Since a cross-validation of the type leave-one-subject-out or leave-n-subjects-out is performed on the available data to quantitatively assess such results, it allows to draw inference on the full population from which the group of participants was drawn, including individuals for which no data was available. As previously pointed out in Kragel et al (2018), the interpretation that follows is straightforward: a significant effect detected with ISPA implies that some information has been identified to be consistent throughout the full population. In more details, this means that the modulations of the multivariate patterns according to the experimental conditions that were the object of the decoding analysis are consistent throughout the population, at least at the resolution offered by the modality used for the acquisition.…”

“…Inter-subject pattern analysis offers a higher detection power to detect weak distributed effects and facilitate the interpretation while the results provided by the information-based approach necessitate further investigation to raise potential ambiguities. Furthermore, because it allows identifying group-wise invariants from functional neuroimaging pat-terns, inter-subject pattern analysis is a tool of choice to identify neuromarkers Gabrieli et al (2015) or brain signatures Kragel et al (2018), making it a versatile scheme for population-wise multivariate analyses.…”

Functional and diffusion MRI reveal the functional and structural basis of infants’ noxious-evoked brain activity

“…However, it is possible to develop a principled approach for noxious-response feature extraction that could decrease this ambiguity and improve biological interpretability. There are now several candidate fMRI neural signatures for distinct components of adult pain and negative affect that could be applied to infant noxious-response data [28][29][30][31][32] , and this is currently being actively researched by our group. Additionally, the reported functional coupling between infant resting-state and stimulus-response activities is currently limited to the nociceptive stimulus modality employed in this study.…”

Volume-wise analysis of fMRI time series allows accurate prediction of language lateralization

“…One reason for this schism might be the framework of voxel-wise analysis, where each voxel is treated as independent of the rest of the brain. Because this approach was not developed with clinical applications in mind, it has been argued that using fMRI as a biomarker requires a shift in perspective (Woo et al, 2017;Kragel et al, 2018). To capitalize on the rich spatial data that fMRI provides even in the individual case, pattern analysis methods try to aggregate information across the whole brain.…”