MULTIVARIATE MACHINE LEARNING APPROACH TO INTEGRATE MULTIMODAL NEUROIMAGING DATASETS
Loading...
Links to Files
Permanent Link
Author/Creator
Author/Creator ORCID
Date
2020-01-20
Type of Work
Department
Computer Science and Electrical Engineering
Program
Computer Science
Citation of Original Publication
Rights
Distribution Rights granted to UMBC by the author.
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
This item may be protected under Title 17 of the U.S. Copyright Law. It is made available by UMBC for non-commercial research and education. For permission to publish or reproduce, please see http://aok.lib.umbc.edu/specoll/repro.php or contact Special Collections at speccoll(at)umbc.edu
Abstract
Non-invasive functional neuroimaging techniques have evolved remarkably over the last decades. Consequently, developing new procedures for more efficient description of neuroimaging data is emerging as a critical requirement. Such procedures, which enable automated extraction of clinically relevant information from neuroimaging data and can display it in an appropriate form, will enable more rapid and efficient evaluations. The main objective of this work is to apply multi-set Canonical Correlation Analysis (mCCA) for two distinct neuroimaging datasets: functional Near Infra-Red Spectroscopy (fNIRS) and ElectroEncephaloGraphy (EEG), to extract an enhanced analysis. However, fNIRS and EEG datasets are enormously different in nature and present challenges in leveraging the information content of both data sets. For example, EEG has excellent temporal resolution (ms range) and scattered spatial resolution, while fNIRS temporal resolution is slow in the range of seconds but its spatial resolution can be more local than EEG. Hence, there is a high possibility that that non-instantaneous relationships can be observed between brain neural and hemodynamic responses. To investigate the relationship between the mentioned datasets, we applied mCCA to find subject associations within each modality and also the correlation between groups of subjects across the two modalities. Utilizing a hybrid EEG/fNIRS correlation system should provide complementary information to better understand and explain the underlying neural/functional brain processes. In addition, we will present fNIRS data analysis case studies relating to a Moral Judgment (MJ) task. Using statistical learning techniques, such as mixed effect modeling, we investigate functional brain activity during personal (emotionally salient) vs impersonal (less emotional and more logical) MJ decision making. We show the proposed methodology can account for the covariance and causality structures present in the signal. This feature makes conventional methods when combined with fNIR more capable than conventional univariate methods alone. Furthermore, using mCCA, we were able to find a relationship between the behavioral datasets and neuroimaging data. Behavioral dataset consists of psychopathic traits measured by a self-report questionnaire called Psychopathic Personality Inventory-Revised (PPI-R). This case study, to our knowledge, is the first to determine the psychopathic core traits most correlated with brain functional activation.