Interpretability of Multivariate Brain Maps in Brain Decoding: Definition and Quantification (2016-03-29T00:00:00.000000Z)

TL;DR

A theoretical definition of interpretability in brain decoding is presented and it is shown that the interpretability of multivariate brain maps can be decomposed into their reproducibility and representativeness and facilitates the development of more effective brain decoding algorithms in the future.

Abstract

Brain decoding is a popular multivariate approach for hypothesis testing in neuroimaging. Linear classifiers are widely employed in the brain decoding paradigm to discriminate among experimental conditions. Then, the derived linear weights are visualized in the form of multivariate brain maps to further study the spatio-temporal patterns of underlying neural activities. It is well known that the brain maps derived from weights of linear classifiers are hard to interpret because of high correlations between predictors, low signal to noise ratios, and the high dimensionality of neuroimaging data. Therefore, improving the interpretability of brain decoding approaches is of primary interest in many neuroimaging studies. Despite extensive studies of this type, at present, there is no formal definition for interpretability of multivariate brain maps. As a consequence, there is no quantitative measure for evaluating the interpretability of different brain decoding methods. In this paper, first, we present a theoretical definition of interpretability in brain decoding; we show that the interpretability of multivariate brain maps can be decomposed into their reproducibility and representativeness. Second, as an application of the proposed definition, we formalize a heuristic method for approximating the interpretability of multivariate brain maps in a binary magnetoencephalography (MEG) decoding scenario. Third, we pro pose to combine the approximated interpretability and the performance of the brain decoding into a new multi-objective criterion for model selection. Our results for the MEG data show that optimizing the hyper-parameters of the regularized linear classifier based on the proposed criterion results in more informative multivariate brain maps. More importantly, the presented definition provides the theoretical background for quantitative evaluation of interpretability, and hence, facilitates the development of more effective brain decoding algorithms in the future.

References116 items

A model explanation system

Classical Statistics and Statistical Learning in Imaging Neuroscience

Resolving Ambiguities of MVPA Using Explicit Models of Representation

A Primer on Pattern-Based Approaches to fMRI: Principles, Pitfalls, and Perspectives

On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation

DECAF: MEG-Based Multimodal Database for Decoding Affective Physiological Responses

A Novel Feature Selection Approach for Analyzing High dimensional Functional MRI Data

A Novel Approach for Stable Selection of Informative Redundant Features from High Dimensional fMRI Data

Learning interpretable classification rules using sequential rowsampling

Linear Maximum Margin Classifier for Learning from Uncertain Data

Causal interpretation rules for encoding and decoding models in neuroimaging

Classification of spectral data using fused lasso logistic regression

Automated model selection in covariance estimation and spatial whitening of MEG and EEG signals

A multi-subject, multi-modal human neuroimaging dataset

Multi-Task Learning for Interpretation of Brain Decoding Models

Practical Applications of Sparse Modeling

How machine learning is shaping cognitive neuroimaging

Dimensionality reduction for the analysis of brain oscillations

Randomized structural sparsity via constrained block subsampling for improved sensitivity of discriminative voxel identification

A Universal and Efficient Method to Compute Maps from Image-Based Prediction Models

What do differences between multi-voxel and univariate analysis mean? How subject-, voxel-, and trial-level variance impact fMRI analysis

Parameter interpretation, regularization and source localization in multivariate linear models

MEG decoding across subjects

On the interpretation of weight vectors of linear models in multivariate neuroimaging

100% Classification Accuracy Considered Harmful: The Normalized Information Transfer Factor Explains the Accuracy Paradox

A Survey of L1 Regression

Decoding magnetoencephalographic rhythmic activity using spectrospatial information

Fast Bootstrapping and Permutation Testing for Assessing Reproducibility and Interpretability of Multivariate fMRI Decoding Models

Identifying Predictive Regions from fMRI with TV-L1 Prior

Analyzing Local Structure in Kernel-Based Learning: Explanation, Complexity, and Reliability Assessment

Interpretable whole-brain prediction analysis with GraphNet

Estimation Stability With Cross-Validation (ESCV)

Privacy-preserving speech processing: cryptographic and string-matching frameworks show promise

Mind reading with regularized multinomial logistic regression

Proceedings of the 29th International Conference on Machine Learning (ICML-12)

Model sparsity and brain pattern interpretation of classification models in neuroimaging

Combining sparseness and smoothness improves classification accuracy and interpretability

Statistical testing in electrophysiological studies.

Beyond Brain Reading: Randomized Sparsity and Clustering to Simultaneously Predict and Identify

Mass univariate analysis of event-related brain potentials/fields I: a critical tutorial review.

Mass univariate analysis of event-related brain potentials/fields II: Simulation studies.

Whole-brain Prediction Analysis with GraphNet

Visual Interpretation of Kernel‐Based Prediction Models

Encoding and decoding in fMRI

Detecting stable distributed patterns of brain activation using Gini contrast

Introduction to machine learning for brain imaging

Model-based feature construction for multivariate decoding

Common component classification: What can we learn from machine learning?

A Parametric Empirical Bayesian Framework for the EEG/MEG Inverse Problem: Generative Models for Multi-Subject and Multi-Modal Integration

Decoding word and category-specific spatiotemporal representations from MEG and EEG

Total Variation Regularization for fMRI-Based Prediction of Behavior

FieldTrip: Open Source Software for Advanced Analysis of MEG, EEG, and Invasive Electrophysiological Data

How to Explain Individual Classification Decisions

Interpreting single trial data using groupwise regularisation

Whole-brain Sparse Penalized Discriminant Analysis for Predicting Choice

Methods and Software for fMRI Analysis of Clinical Subjects

Mutual Information Based Metric for Evaluation of fMRI Data Processing Approaches

A Survey of Uncertain Data Algorithms and Applications

Attention modulations of posterior alpha as a control signal for two-dimensional brain–computer interfaces

Structured Variable Selection with Sparsity-Inducing Norms

Machine learning classifiers and fMRI: A tutorial overview

Prediction and interpretation of distributed neural activity with sparse models

Multi-objective Management in Freight Logistics: Increasing Capacity, Service Level and Safety with Optimization Algorithms

Interpretable Classifiers for fMRI Improve Prediction of Purchases

Hand Movement Direction Decoded from MEG and EEG

Nonparametric statistical testing of EEG- and MEG-data

Soul, mind, brain: Greek philosophy and the birth of neuroscience

Beyond mind-reading: multi-voxel pattern analysis of fMRI data

Neuroimaging: Decoding mental states from brain activity in humans

Complete functional characterization of sensory neurons by system identification.

Estimation of Dependences Based on Empirical Data

Information-based functional brain mapping.

Model selection and estimation in regression with grouped variables

Addendum: Regularization and variable selection via the elastic net

Sparsity and smoothness via the fused lasso

Support Vector Classification with Input Data Uncertainty

Bias-Variance Analysis of Support Vector Machines for the Development of SVM-Based Ensemble Methods

Learning to Decode Cognitive States from Brain Images

Survey of multi-objective optimization methods for engineering

Single-trial detection in EEG and MEG: Keeping it linear

Functional magnetic resonance imaging (fMRI) “brain reading”: detecting and classifying distributed patterns of fMRI activity in human visual cortex

Brain–computer interfaces: a review

Brain–computer interfaces for communication and control

On the mathematical foundations of learning

Distributed and Overlapping Representations of Faces and Objects in Ventral Temporal Cortex

A Unified Bias-Variance Decomposition for Zero-One and Squared Loss

Predicting the recognition of natural scenes from single trial MEG recordings of brain activity

An Efficient Method To Estimate Bagging's Generalization Error

Nonlinear versus Linear Models in Functional Neuroimaging: Learning Curves and Generalization Crossover

Electrophysiological Studies of Face Perception in Humans

Statistical methods of estimation and inference for functional MR image analysis

A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection

Multi-objective Optimization

0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not

Classification accuracy as a proxy for 847 two sample testing , arXiv preprint arXiv : 1602

Brain-Computer Interfaces

High-Dimensional Sparse Structured Input-Output Models, with Applications to GWAS

Stability and Reproducibility in fMRI Analysis

Novel noise reduction meth- 897 ods

100

Bernoulli

101

Author manuscript, published in "International Conference on Machine Learning (2012)" Small-sample brain mapping: sparse recovery on spatially correlated designs with randomization and clustering

102

Making machine learning models interpretable

103

On the Interpretability of Linear Multivariate Neuroimaging Analyses : Filters , Patterns and their Relationship

104

Learning and Interpretation in Neuroimaging

105

Bootstrap Methods: Another Look at the Jackknife

106

Classification methods for ongoing EEG and MEG signals.

107

Stability and Generalization

108

Random Forests

109

The Elements of Statistical Learning: Data Mining, Inference, and Prediction

110

The Nature of Statistical Learning Theory

111

An Eecient Method to Estimate Bagging's Generalization Error

112

Regression Shrinkage and Selection via the Lasso

113

Bias, Variance and Prediction Error for Classification Rules

114

Electrophysiology of Mind: Event-Related Brain Potentials and Cognition

115

ficial Neural Networks, Computational Intelligence and Machine Learning

116

Proceedings of the 20th European Symposium on Arti