Analysis Of Brain Imaging Data Research Articles

On Being Transhuman: Commercial BCIs and the Quest for Autonomy

Facebook’s catchphrase “What’s on your mind”, prompting the user to share thoughts with their digital social circles, has gained a new, literal meaning in recent years: targeted advertising, fake news and computational propaganda all being examples of mental ma-nipulation exerted for profit or for power through harnessing AI at scale for the purposes of online profiling. In most cases, this in-volves an elaborate interpretation of one’s digital footprint: the huge amount of data that is generated by our daily online and offline in-teractions and which defines our behaviour. This chapter takes a slightly different approach and seeks to explore the use of AI to re-trieve, analyse and predict data that has not been externalised, yet which most defines us: brain data. There has never been a more promising time in history for delineating the contours of human thought: Public and privately funded projects studying the human brain have produced a high vol-ume of scientific papers and findings in the last few decades, which more often than not are sensationalised in the news. The ambitious plan to explain the mysteries of the human brain has not fully materialised, however ambition drives profit, and therefore the idea of using AI to decode the human brain has been a fast growing commercial venture for many tech giants, who have been investing heavily in corporate R & D neurotech related projects. The chapter proceeds in four parts: Part 1 offers a historical overview of “mind reading” techniques, building up some context as to how the neurotech market boomed and started employing AI to unravel the mysteries of the human brain beyond the clinical sphere. This is then followed by a techno-legal evaluation of the monitoring, collection and analysis of brain imaging data from the use of com-mercial BCIs in Part 2. Building further on this, Part 3 explores the scope for user empowerment and agency in commercial BCI. This will lead to the main argument put forth in Part 4, namely the fact that commercial BCI constitutes a special case that seems to fall through the cracks of robust data protection frameworks, such as the GDPR. In conclusion, the chapter highlights the need for data protection laws to reflect the conceptual redefinition of autonomy in the light of AI- driven pervasive neurotechnologies.

An algorithm for learning shape and appearance models without annotations.

This paper presents a framework for automatically learning shape and appearance models for medical (and certain other) images. The algorithm was developed with the aim of eventually enabling distributed privacy-preserving analysis of brain image data, such that shared information (shape and appearance basis functions) may be passed across sites, whereas latent variables that encode individual images remain secure within each site. These latent variables are proposed as features for privacy-preserving data mining applications. The approach is demonstrated qualitatively on the KDEF dataset of 2D face images, showing that it can align images that traditionally require shape and appearance models trained using manually annotated data (manually defined landmarks etc.). It is applied to the MNIST dataset of handwritten digits to show its potential for machine learning applications, particularly when training data is limited. The model is able to handle "missing data", which allows it to be cross-validated according to how well it can predict left-out voxels. The suitability of the derived features for classifying individuals into patient groups was assessed by applying it to a dataset of over 1900 segmented T1-weighted MR images, which included images from the COBRE and ABIDE datasets.

Blind fMRI source unmixing via higher-order tensor decompositions

BackgroundThe growing interest in neuroimaging technologies generates a massive amount of biomedical data of high dimensionality. Tensor-based analysis of brain imaging data has been recognized as an effective analysis that exploits its inherent multi-way nature. In particular, the advantages of tensorial over matrix-based methods have previously been demonstrated in the context of functional magnetic resonance imaging (fMRI) source localization. However, such methods can also become ineffective in realistic challenging scenarios, involving, e.g., strong noise and/or significant overlap among the activated regions. Moreover, they commonly rely on the assumption of an underlying multilinear model generating the data. New methodThis paper aims at investigating the possible gains from exploiting the 4-dimensional nature of the brain images, through a higher-order tensorization of the fMRI signal, and the use of less restrictive generative models. In this context, the higher-order block term decomposition (BTD) and the PARAFAC2 tensor models are considered for the first time in fMRI blind source separation. A novel PARAFAC2-like extension of BTD (BTD2) is also proposed, aiming at combining the effectiveness of BTD in handling strong instances of noise and the potential of PARAFAC2 to cope with datasets that do not follow the strict multilinear assumption. Comparison with existing methodsThe methods were tested using both synthetic and real data and compared with state of the art methods. ConclusionsThe simulation results demonstrate the effectiveness of BTD and BTD2 for challenging scenarios (presence of noise, spatial overlap among activation regions and inter-subject variability in the haemodynamic response function (HRF)).

Spatial functional principal component analysis with applications to brain image data

This paper considers a fast and effective algorithm for conducting functional principal component analysis with multivariate factors. Compared with the univariate case, our approach could be more powerful in revealing spatial connections or extracting important features in images. To facilitate fast computation, we connect singular value decomposition with penalized smoothing and avoid estimating a covariance operator in very high dimension. Under regularity assumptions, the results indicate that we may enjoy the optimal convergence rate by employing the smoothness assumption inherent to functional objects. We apply our method to the analysis of brain image data. Our extracted factors provide excellent recovery of the risk related regions of interest in the human brain and the estimated loadings are very informative in revealing individual risk attitude.

Decentralized Analysis of Brain Imaging Data: Voxel-Based Morphometry and Dynamic Functional Network Connectivity

In the field of neuroimaging, there is a growing interest in developing collaborative frameworks that enable researchers to address challenging questions about the human brain by leveraging data across multiple sites all over the world. Additionally, efforts are also being directed at developing algorithms that enable collaborative analysis and feature learning from multiple sites without requiring the often large data to be centrally located. In this paper, we propose two new decentralized algorithms: (1) A decentralized regression algorithm for performing a voxel-based morphometry analysis on structural magnetic resonance imaging (MRI) data and, (2) A decentralized dynamic functional network connectivity algorithm which includes decentralized group ICA and sliding-window analysis of functional MRI data. We compare results against those obtained from their pooled (or centralized) counterparts on the same data i.e., as if they are at one site. Results produced by the decentralized algorithms are similar to the pooled-case and showcase the potential of performing multi-voxel and multivariate analyses of data located at multiple sites. Such approaches enable many more collaborative and comparative analysis in the context of large-scale neuroimaging studies.

A Graph Signal Processing Perspective on Functional Brain Imaging

Modern neuroimaging techniques provide us with unique views on brain structure and function; i.e., how the brain is wired, and where and when activity takes place. Data acquired using these techniques can be analyzed in terms of its network structure to reveal organizing principles at the systems level. Graph representations are versatile models where nodes are associated to brain regions and edges to structural or functional connections. Structural graphs model neural pathways in white matter, which are the anatomical backbone between regions. Functional graphs are built based on functional connectivity, which is a pairwise measure of statistical interdependency between pairs of regional activity traces. Therefore, most research to date has focused on analyzing these graphs reflecting structure or function. Graph signal processing (GSP) is an emerging area of research where signals recorded at the nodes of the graph are studied atop the underlying graph structure. An increasing number of fundamental operations have been generalized to the graph setting, allowing to analyze the signals from a new viewpoint. Here, we review GSP for brain imaging data and discuss their potential to integrate brain structure, contained in the graph itself, with brain function, residing in the graph signals. We review how brain activity can be meaningfully filtered based on concepts of spectral modes derived from brain structure. We also derive other operations such as surrogate data generation or decompositions informed by cognitive systems. In sum, GSP offers a novel framework for the analysis of brain imaging data.

Application of Graph Theory to Assess Static and Dynamic Brain Connectivity: Approaches for Building Brain Graphs.

Human brain connectivity is complex. Graph theory based analysis has become a powerful and popular approach for analyzing brain imaging data, largely because of its potential to quantitatively illuminate the networks, the static architecture in structure and function, the organization of dynamic behavior over time, and disease related brain changes. The first step in creating brain graphs is to define the nodes and edges connecting them. We review a number of approaches for defining brain nodes including fixed versus data-driven nodes. Expanding the narrow view of most studies which focus on static and/or single modality brain connectivity, we also survey advanced approaches and their performances in building dynamic and multi-modal brain graphs. We show results from both simulated and real data from healthy controls and patients with mental illnesse. We outline the advantages and challenges of these various techniques. By summarizing and inspecting recent studies which analyzed brain imaging data based on graph theory, this article provides a guide for developing new powerful tools to explore complex brain networks.

205. Deep Learning Approaches to Unimodal and Multimodal Analysis of Brain Imaging Data With Applications to Mental Illness

205. Deep Learning Approaches to Unimodal and Multimodal Analysis of Brain Imaging Data With Applications to Mental Illness

Evolution of Abstraction Across Layers in Deep Learning Neural Networks

Deep learning neural networks produce excellent results in various pattern recognition tasks. It is of great practical importance to answer some open questions regarding model design and parameterization, and to understand how input data are converted into meaningful knowledge at the output. The layer-by-layer evolution of the abstraction level has been proposed previously as a quantitative measure to describe the emergence of knowledge in the network. In this work we systematically evaluate the abstraction level for a variety of image datasets. We observe that there is a general tendency of increasing abstraction from input to output with the exception of a drop of abstraction at some ReLu and Pooling layers. The abstraction level is relatively low and does not change significantly in the first few layers following the input, while it fluctuates around some high saturation value at the layers preceding the output. Finally, the layer-by-layer change in abstraction is not normally distributed, rather it approximates an exponential distribution. These results point to salient local features of deep layers impacting overall (global) classification performance. We compare the results extracted from deep learning neural networks performing image processing tasks with the results obtained by analyzing brain imaging data. Our conclusions may be helpful in future designs of more efficient, compact deep learning neural networks.

Pattern Analyses Reveal Separate Experience-Based Fear Memories in the Human Right Amygdala.

Learning fear via the experience of contingencies between a conditioned stimulus (CS) and an aversive unconditioned stimulus (US) is often assumed to be fundamentally different from learning fear via instructions. An open question is whether fear-related brain areas respond differently to experienced CS-US contingencies than to merely instructed CS-US contingencies. Here, we contrasted two experimental conditions where subjects were instructed to expect the same CS-US contingencies while only one condition was characterized by prior experience with the CS-US contingency. Using multivoxel pattern analysis of fMRI data, we found CS-related neural activation patterns in the right amygdala (but not in other fear-related regions) that dissociated between whether a CS-US contingency had been instructed and experienced versus merely instructed. A second experiment further corroborated this finding by showing a category-independent neural response to instructed and experienced, but not merely instructed, CS presentations in the human right amygdala. Together, these findings are in line with previous studies showing that verbal fear instructions have a strong impact on both brain and behavior. However, even in the face of fear instructions, the human right amygdala still shows a separable neural pattern response to experience-based fear contingencies.SIGNIFICANCE STATEMENT In our study, we addressed a fundamental problem of the science of human fear learning and memory, namely whether fear learning via experience in humans relies on a neural pathway that can be separated from fear learning via verbal information. Using two new procedures and recent advances in the analysis of brain imaging data, we localized purely experience-based fear processing and memory in the right amygdala, thereby making a direct link between human and animal research.

5 Computational neuropsychiatry

Professor of Neuropsychiatry and Director of the Wellcome Trust Centre for Neuroimaging, UCL, London. Ray Dolan is Mary Kinross Professor of Neuropsychiatry and Director of the Wellcome Trust Centre for Neuroimaging, at UCL. Prof Dolan’s research addresses the neurobiology of emotion and decision making, how emotion impacts on cognition and its aberrant expression in disease. He has published over 500 peer reviewed papers is among the most cited scientist in the world in the field of Neuroscience and Behaviour. He has received numerous awards including the Alexander von Humboldt Research Award (2004), the Minerva Foundation Golden Brain Award (2006), the International Max Planck Research Award (2007) and the Zulch Prize (2013). Between 2010–2014 he has been Visiting Einstein Fellow to the Humboldt University, Berlin. He was elected Fellow of the Academy of Medical Sciences (FMedSci) in 2000, Fellow of the Royal Society (FRS) in 2010 and an External Member of the Max Planck Society (MPS) in 2012. He recently established the first Centre for Computational Psychiatry at UCL, a joint venture with the Max Planck Society (MPS). Computational psychiatry encapsulates a set of approaches to the study of psychiatric conditions that are rooted in computer science. Two broad approaches can be identified. The first of these is an atheoretical and with an emphasis that is data driven. This usually involves the application of machine learning (ML) methods in an attempt to refine disease classification, predict treatment response and improve treatment outcomes. This approach can be applied to a wide class of data, including the analysis of brain imaging data. The second approach is theoretical and grounded within developments that derive from theoretical neuroscience. My focus in this talk will be entirely on the latter where I will provide examples of its utility in deepening our understanding of (i) adaptive learning and risk behaviour (ii) providing a quantitative basis for measuring and predicting subjective states (iii) providing neurobiologically inspired models that can be used to inform mechanistic interpretations of large scale data. These examples will serve to highlight how in computational psychiatry can open new avenues of investigation that provide mechanistic insights into disease processes with implications for classification and treatment. Computational psychiatry also enables an easier dialogue between advances in neuroscience and the challenges of providing a biologically informed understating of psychopathology.

Spatial Functional Principal Component Analysis with Applications to Brain Image Data

This paper considers a fast and effective algorithm for conducting functional principle component analysis with multivariate factors. Compared with the univariate case, our approach could be more powerful in revealing spatial connections or extracting important features in images. To facilitate fast computation, we connect Singular Value Decomposition with penalized smoothing and avoid estimating a huge dimensional covariance operator. Under regularity assumptions, the results indicate that we may enjoy the optimal convergence rate by employing the smoothness assumption inherent to functional objects. We apply our method on the analysis of brain image data. Our extracted factors provide excellent recovery of the risk related regions of interests in human brain and the estimated loadings are very informative in revealing the individual risk attitude.

Feature Selection for Microarray Data using WGCNA Based Fuzzy Forest in Map Reduce Paradigm

Objectives: The feature selection is one category of principally used information analysis algorithms on microarray information or any related to high dimensional information. The goal of the feature selection algorithms is to separate out a little set of informative options that best explains experimental variations. This work really investigates the feature selection drawback for microarray information with tiny samples and variant correlation. Most existing algorithms sometimes need expensive machine effort, particularly beneath thousands of cistron (Gene) conditions. Usually citron (Gene) selection methodology searches for associate best or close to best set of genes with relevance a given analysis. Methods: The main objective of this project is to effectively choose the foremost informative options from microarray data, whereas creating the machine expenses reasonable. This can be achieved by proposing Fuzzy Forest using Weighted Gene Correlation Network Analysis (WGCNA) that makes use of interaction between the features (or Genes). Necessary representative features (or Genes) selection measure designated from every enriched feature partition to make the reduced gene area. Findings: Finally, by shaping a correct regression context, the planned methodology are often simply implement to utilize the MapReduce paradigm, that considerably reduces machine load and additionally resulting in lower prediction error rates (OOB) with variable importance compared to different existing approaches. Thus, it is necessary to know the performance of random forest with microarray data and its potential use for gene selection. Applications: Some the major applications of WGCNA based Fuzzy Forest are genomic data analysis (including microarray data), neuroscience, bioinformatics, DNA methylation data analysis (16s rRNA gene sequencing), cancer, yeast genetics analysis, analysis of brain imaging data (functional MRI data analysis).

Regional selection of the brain size regulating gene CASC5 provides new insight into human brain evolution.

Human evolution is marked by a continued enlargement of the brain. Previous studies on human brain evolution focused on identifying sequence divergences of brain size regulating genes between humans and nonhuman primates. However, the evolutionary pattern of the brain size regulating genes during recent human evolution is largely unknown. We conducted a comprehensive analysis of the brain size regulating gene CASC5 and found that in recent human evolution, CASC5 has accumulated many modern human specific amino acid changes, including two fixed changes and six polymorphic changes. Among human populations, 4 of the 6 amino acid polymorphic sites have high frequencies of derived alleles in East Asians, but are rare in Europeans and Africans. We proved that this between-population allelic divergence was caused by regional Darwinian positive selection in East Asians. Further analysis of brain image data of Han Chinese showed significant associations of the amino acid polymorphic sites with gray matter volume. Hence, CASC5 may contribute to the morphological and structural changes of the human brain during recent evolution. The observed between-population divergence of CASC5 variants was driven by natural selection that tends to favor a larger gray matter volume in East Asians.

A Projection free method for Generalized Eigenvalue Problem with a nonsmooth Regularizer.

Eigenvalue problems are ubiquitous in computer vision, covering a very broad spectrum of applications ranging from estimation problems in multi-view geometry to image segmentation. Few other linear algebra problems have a more mature set of numerical routines available and many computer vision libraries leverage such tools extensively. However, the ability to call the underlying solver only as a "black box" can often become restrictive. Many 'human in the loop' settings in vision frequently exploit supervision from an expert, to the extent that the user can be considered a subroutine in the overall system. In other cases, there is additional domain knowledge, side or even partial information that one may want to incorporate within the formulation. In general, regularizing a (generalized) eigenvalue problem with such side information remains difficult. Motivated by these needs, this paper presents an optimization scheme to solve generalized eigenvalue problems (GEP) involving a (nonsmooth) regularizer. We start from an alternative formulation of GEP where the feasibility set of the model involves the Stiefel manifold. The core of this paper presents an end to end stochastic optimization scheme for the resultant problem. We show how this general algorithm enables improved statistical analysis of brain imaging data where the regularizer is derived from other 'views' of the disease pathology, involving clinical measurements and other image-derived representations.

Subcortical brain volume abnormalities in 2028 individuals with schizophrenia and 2540 healthy controls via the ENIGMA consortium.

The profile of brain structural abnormalities in schizophrenia is still not fully understood, despite decades of research using brain scans. To validate a prospective meta-analysis approach to analyzing multicenter neuroimaging data, we analyzed brain MRI scans from 2028 schizophrenia patients and 2540 healthy controls, assessed with standardized methods at 15 centers worldwide. We identified subcortical brain volumes that differentiated patients from controls, and ranked them according to their effect sizes. Compared with healthy controls, patients with schizophrenia had smaller hippocampus (Cohen's d=−0.46), amygdala (d=−0.31), thalamus (d=−0.31), accumbens (d=−0.25) and intracranial volumes (d=−0.12), as well as larger pallidum (d=0.21) and lateral ventricle volumes (d=0.37). Putamen and pallidum volume augmentations were positively associated with duration of illness and hippocampal deficits scaled with the proportion of unmedicated patients. Worldwide cooperative analyses of brain imaging data support a profile of subcortical abnormalities in schizophrenia, which is consistent with that based on traditional meta-analytic approaches. This first ENIGMA Schizophrenia Working Group study validates that collaborative data analyses can readily be used across brain phenotypes and disorders and encourages analysis and data sharing efforts to further our understanding of severe mental illness.

Full correlation matrix analysis (FCMA): An unbiased method for task-related functional connectivity

BackgroundThe analysis of brain imaging data often requires simplifying assumptions because exhaustive analyses are computationally intractable. Standard univariate and multivariate analyses of brain activity ignore interactions between regions and analyses of interactions (functional connectivity) reduce the computational challenge by using seed regions of interest or brain parcellations. New methodTo meet this challenge, we developed full correlation matrix analysis (FCMA), which leverages and optimizes algorithms from parallel computing and machine learning to efficiently analyze the pairwise correlations of all voxels in the brain during different cognitive tasks, with the goal of identifying task-related interactions in an unbiased manner. ResultsWhen applied to a localizer dataset on a small compute cluster, FCMA accelerated a naive, serial approach by four orders of magnitude, reducing running time from two years to one hour. In addition to this performance gain, FCMA emphasized different brain areas than existing methods. In particular, beyond replicating known category selectivity in visual cortex, FCMA also revealed a region of medial prefrontal cortex whose selectivity derived from differential patterns of functional connectivity across categories. Comparison with existing method(s)For benchmarking, we started with a naive approach and progressively built up to the complete FCMA procedure by adding optimized classifier algorithms, multi-threaded parallelism, and multi-node parallelism. To evaluate what can be learned with FCMA, we compared it against multivariate pattern analysis of activity and seed-based analysis of functional connectivity. ConclusionsFCMA demonstrates how advances in computer science can alleviate computational bottlenecks in neuroscience. We have released a software toolbox to help others evaluate FCMA.

Statistical inferences under the Null hypothesis: common mistakes and pitfalls in neuroimaging studies.

Published studies using functional and structural MRI include many errors in the way data are analyzed and conclusions reported. This was observed when working on a comprehensive review of the neural bases of synesthesia, but these errors are probably endemic to neuroimaging studies. All studies reviewed had based their conclusions using Null Hypothesis Significance Tests (NHST). NHST have yet been criticized since their inception because they are more appropriate for taking decisions related to a Null hypothesis (like in manufacturing) than for making inferences about behavioral and neuronal processes. Here I focus on a few key problems of NHST related to brain imaging techniques, and explain why or when we should not rely on “significance” tests. I also observed that, often, the ill-posed logic of NHST was even not correctly applied, and describe what I identified as common mistakes or at least problematic practices in published papers, in light of what could be considered as the very basics of statistical inference. MRI statistics also involve much more complex issues than standard statistical inference. Analysis pipelines vary a lot between studies, even for those using the same software, and there is no consensus which pipeline is the best. I propose a synthetic view of the logic behind the possible methodological choices, and warn against the usage and interpretation of two statistical methods popular in brain imaging studies, the false discovery rate (FDR) procedure and permutation tests. I suggest that current models for the analysis of brain imaging data suffer from serious limitations and call for a revision taking into account the “new statistics” (confidence intervals) logic.

Including signal intensity increases the performance of blind source separation on brain imaging data.

When analyzing brain imaging data, blind source separation (BSS) techniques critically depend on the level of dimensional reduction. If the reduction level is too slight, the BSS model would be overfitted and become unavailable. Thus, the reduction level must be set relatively heavy. This approach risks discarding useful information and crucially limits the performance of BSS techniques. In this study, a new BSS method that can work well even at a slight reduction level is presented. We proposed the concept of "signal intensity" which measures the significance of the source. Only picking the sources with significant intensity, the new method can avoid the overfitted solutions which are nonexistent artifacts. This approach enables the reduction level to be set slight and retains more useful dimensions in the preliminary reduction. Comparisons between the new and conventional algorithms were performed on both simulated and real data.

SELF-SUPERVISED MRI TISSUE SEGMENTATION BY DISCRIMINATIVE CLUSTERING

The study of brain lesions can benefit from a clear identification of transitions between healthy and pathological tissues, through the analysis of brain imaging data. Current signal processing methods, able to address these issues, often rely on strong prior information. In this article, a new method for tissue segmentation is proposed. It is based on a discriminative strategy, in a self-supervised machine learning approach. This method avoids the use of prior information, which makes it very versatile, and able to cope with different tissue types. It also returns tissue probabilities for each voxel, crucial for a good characterization of the evolution of brain lesions. Simulated as well as real benchmark data were used to validate the accuracy of the method and compare it against other segmentation algorithms.