Decoding Neural Representational Spaces Using Multivariate Pattern Analysis

Abstract

A major challenge for systems neuroscience is to break the neural code. Computational algorithms for encoding information into neural activity and extracting information from measured activity afford understanding of how percepts, memories, thought, and knowledge are represented in patterns of brain activity. The past decade and a half has seen significant advances in the development of methods for decoding human neural activity, such as multivariate pattern classification, representational similarity analysis, hyperalignment, and stimulus-model-based encoding and decoding. This article reviews these advances and integrates neural decoding methods into a common framework organized around the concept of high-dimensional representational spaces.

A statistical framework for neuroimaging data analysis based on mutual information estimated via a gaussian copula

“…7A). The use of such approaches within neuroimaging is growing, with development and application of techniques such as representational similarity analysis (RSA) [Kriegeskorte and Kievit, 2013], and the use of supervised classification algorithms together with cross‐validation [Hastie et al, 2001], often referred to as a decoding [Haxby et al, 2014; King and Dehaene, 2014; Quian Quiroga and Panzeri, 2009]. Our new GCMI estimator allows us to address these issues within the unified framework of information theory.…”

Brain imaging tests for chronic pain: medical, legal and ethical issues and recommendations

“…Features can relate to the location, magnitude or spatial extent of activation, or functional connectivity in a network or between specific brain areas; these variables are correlated with a behavioural measure, such as a reported pain experience (BOX 3). Multivariate approaches integrate multiple features of brain imaging data into an integrated predictive model 3,4,59,60 . Machine learning and statistical techniques are often used to identify patterns in these data, and are optimized to jointly predict patient status, the experience of pain, analgesia, and other outcomes.…”

“…When applied to functional brain images, machine learning can be used to detect response patterns (for example, intensity and spatial distribution of functional MRI (fMRI) signals or spatial-temporal patterns of EEG signals) associated with a given experimental variable (for example, the intensity of pain perception) 61,126 . Multivariate pattern analysis (MVPA) is a machine learning technique in which a pattern classifier identifies the fMRI responses elicited by different stimuli 3,4,59,60 . MVPA is different from conventional univariate fMRI analysis approaches, which use a general linear model to detect average regional activity and consider a single voxel at a time within a given brain region (see figure, top right).…”

“…Karen D. Davis 1,2,3 , Herta Flor 4 , Henry T. Greely 5 , Gian Domenico Iannetti 6 , Sean Mackey 7 , Markus Ploner 8 , Amanda Pustilnik 9,10 , Irene Tracey 11 , Rolf-Detlef Treede 12 and Tor D. Wager 13,14 Abstract | Chronic pain is the greatest source of disability globally and claims related to chronic pain feature in many insurance and medico-legal cases. Brain imaging (for example, functional MRI, PET, EEG and magnetoencephalography) is widely considered to have potential for diagnosis, prognostication, and prediction of treatment outcome in patients with chronic pain.…”

“…Chronic pain is defined as pain that is present every day for >3 months (or is present on ≥50% of days for 6 months) or beyond the expected period of healing and does not have the warning function that acute pain does [1][2][3][4] . Chronic pain is associated with enormous personal and societal costs; individuals with chronic pain often have a reduced quality of life and unmet therapeutic needs, and society is struggling to cope with the large numbers of people with this condition.…”