The goal of the present study was to apply deep learning algorithms to identify autism spectrum disorder (ASD) patients from large brain imaging dataset, based solely on the patients brain activation patterns. We investigated ASD patients brain imaging data from a world-wide multi-site database known as ABIDE (Autism Brain Imaging Data Exchange). ASD is a brain-based disorder characterized by social deficits and repetitive behaviors. According to recent Centers for Disease Control data, ASD affects one in 68 children in the United States. We investigated patterns of functional connectivity that objectively identify ASD participants from functional brain imaging data, and attempted to unveil the neural patterns that emerged from the classification. The results improved the state-of-the-art by achieving 70% accuracy in identification of ASD versus control patients in the dataset. The patterns that emerged from the classification show an anticorrelation of brain function between anterior and posterior areas of the brain; the anticorrelation corroborates current empirical evidence of anterior-posterior disruption in brain connectivity in ASD. We present the results and identify the areas of the brain that contributed most to differentiating ASD from typically developing controls as per our deep learning model.
There has been a growing interest in the neuroimaging community regarding resting state data (i.e., passive mental activity) and the subsequent activation of the so-called default mode network (DMN). Although this network was originally characterized by a pattern of deactivation during active cognitive states, more recent applications of data-driven techniques such as independent component analysis (ICA) have permitted the analysis of brain activation during extended periods of truly passive mental activity. However, ICA requires the resultant components to be evaluated for “goodness of fit” via either human raters or more automated techniques. To our knowledge, an investigation on the reliability of either technique in determining the component that best corresponds to default-mode activity has not been performed. Moreover, it is not clear how automated techniques, which are necessarily dependent upon a template mask, are affected by the structures used to compose the mask. The current study investigated both interrater (human-human) reliability and intermethod (human-machine) reliability for determining DMN activation in 42 healthy controls. Results indicated that near perfect interrater reliability was achieved, whereas intermethod reliability was only within the moderate range. The latter was significantly improved via a weighted combination of the anterior and posterior cingulate nodes of the DMN. Implications for fully automating the component selection process are discussed.
Changes in the default mode network (DMN) have been linked to multiple neurological disorders including schizophrenia. The anticorrelated relationship the DMN shares with task-related networks permits the quantification of this network both during task (task-induced deactivations: TID) and during periods of passive mental activity (extended rest). However, the effects of different methodologies (TID vs. extended rest) for quantifying the DMN in the same clinical population are currently not well understood. Moreover, several different analytic techniques, including independent component analyses (ICA) and seed-based correlation analyses exist for examining functional connectivity during extended resting states. The current study compared both methodologies and analytic techniques in a group of patients with schizophrenia (SP) and matched healthy controls (HC). Results indicated that TID analyses, ICA, and seed-based correlation all consistently identified the midline (anterior and posterior cingulate gyrus) and lateral parietal cortex as core regions of the DMN, as well as more variable involvement of temporal lobe structures. In addition, SP exhibited increased deactivation during task, as well as decreased functional connectivity with frontal regions and increased connectivity with posterior and subcortical areas during periods of extended rest. The increased posterior and reduced anterior connectivity may partially explain some of the cognitive dysfunction and clinical symptoms that are frequently associated with schizophrenia.
One of the most consistent electrophysiological deficits reported in the schizophrenia literature is a failure to inhibit, or properly gate, the neuronal response to the second stimulus of an identical pair (i.e., sensory gating). Although animal and invasive human studies have consistently implicated the auditory cortex, prefrontal cortex and hippocampus in mediating the sensory gating response, localized activation in these structures has not always been reported during non-invasive imaging modalities. In the current experiment, event-related FMRI and a variant of the traditional gating paradigm were utilized to examine how the gating network differentially responded to the processing of pairs of identical and non-identical tones. Two single-tone conditions were also presented so that they could be used to estimate the HRF for paired stimuli, reconstructed based on actual hemodynamic responses, to serve as a control non-gating condition. Results supported an emerging theory that the gating response for both paired-tones conditions was primarily mediated by auditory and prefrontal cortex, with potential contributions from the thalamus. Results also indicated that the left auditory cortex may play a preferential role in determining the stimuli that should be inhibited (gated) or receive further processing due to novelty of information. In contrast, there was no evidence of hippocampal involvement, suggesting that future work is needed to determine what role it may play in the gating response.
With resting-state functional MRI (rs-fMRI) there are a variety of post-processing methods that can be used to quantify the human brain connectome. However, there is also a choice of which preprocessing steps will be used prior to calculating the functional connectivity of the brain. In this manuscript, we have tested seven different preprocessing schemes and assessed the reliability between and reproducibility within the various strategies by means of graph theoretical measures. Different preprocessing schemes were tested on a publicly available dataset, which includes rs-fMRI data of healthy controls. The brain was parcellated into 190 nodes and four graph theoretical (GT) measures were calculated; global efficiency (GEFF), characteristic path length (CPL), average clustering coefficient (ACC), and average local efficiency (ALE). Our findings indicate that results can significantly differ based on which preprocessing steps are selected. We also found dependence between motion and GT measurements in most preprocessing strategies. We conclude that by using censoring based on outliers within the functional time-series as a processing, results indicate an increase in reliability of GT measurements with a reduction of the dependency of head motion.
The segmentation of the hippocampus in Magnetic Resonance Imaging (MRI) has been an important procedure to diagnose and monitor several clinical situations. The precise delineation of the borders of this brain structure makes it possible to obtain a measure of the volume and estimate its shape, which can be used to diagnose some diseases, such as Alzheimer's disease, schizophrenia and epilepsy. As the manual segmentation procedure in three-dimensional images is highly time consuming and the reproducibility is low, automated methods introduce substantial gains. On the other hand, the implementation of those methods is a challenge because of the low contrast of this structure in relation to the neighboring areas of the brain. Within this context, this research presents a review of the evolution of automatized methods for the segmentation of the hippocampus in MRI. Many proposed methods for segmentation of the hippocampus have been published in leading journals in the medical image processing area. This paper describes these methods presenting the techniques used and quantitatively comparing the methods based on Dice Similarity Coefficient. Finally, we present an evaluation of those methods considering the degree of user intervention, computational cost, segmentation accuracy and feasibility of application in a clinical routine.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.