Repository logo

Browsing by Author "Rojas, Don, committee member"

Now showing 1 - 7 of 7
  • Thumbnail Image
    ItemOpen Access
    A comparison of tri-polar concentric ring electrodes to disc electrodes for decoding real and imaginary finger movements
    (Colorado State University. Libraries, 2019) Alzahrani, Saleh Ibrahim, author; Anderson, Charles W., advisor; Vigh, Jozsef, committee member; Rojas, Don, committee member; Abdel-Ghany, Salah, committee member
    The electroencephalogram (EEG) is broadly used for diagnosis of brain diseases and research of brain activities. Although the EEG provides a good temporal resolution, it suffers from poor spatial resolution due to the blurring effects of volume conduction and signal-to-noise ratio. Many efforts have been devoted to the development of novel methods that can increase the EEG spatial resolution. The surface Laplacian, which is the second derivative of the surface potential, has been applied to EEG to improve the spatial resolution. Tri-polar concentric ring electrodes (TCREs) have been shown to estimate the surface Laplacian automatically with better spatial resolution than conventional disc electrodes. The aim of this research is to study how well the TCREs can be used to acquire EEG signals to decode real and imaginary finger movements. These EEG signals will be then translated into finger movements commands. We also compare the feasibility of discriminating finger movements from one hand using EEG recorded from TCREs and conventional disc electrodes. Furthermore, we evaluated two movement-related features, temporal EEG data and spectral features, in discriminating individual finger from one hand using non-invasive EEG. To do so, movement-related potentials (MRPs) are measured and analyzed from four TCREs and conventional disc electrodes while 13 subjects performed either motor execution or motor imagery of individual finger movements. The tri-polar-EEG (tEEG) and conventional EEG (cEEG) were recorded from electrodes placed according to the 10-20 International Electrode Positioning System over the motor cortex. Our results show that the TCREs achieved higher spatial resolution than conventional disc electrodes. Moreover, the results show that signals from TCREs generated higher decoding accuracy compared to signals from conventional disc electrodes. The average decoding accuracy of five-class classification for all subjects was of 70.04 ± 7.68% when we used temporal EEG data as feature and classified it using Artificial Neural Networks (ANNs) classifier. In addition, the results show that the TCRE EEG (tEEG) provides approximately a four times enhancement in the signal-to-noise ratio (SNR) compared to disc electrode signals. We also evaluated the interdependency level between neighboring electrodes from tri-polar, disc, and disc with Hjorth's Laplacian method in time and frequency domains by calculating the mutual information (MI) and coherence. The MRP signals recorded with the TCRE system have significantly less mutual information (MI) between electrodes than the conventional disc electrode system and disc electrodes with Hjorth's Laplacian method. Also, the results show that the mean coherence between neighboring tri-polar electrodes was found to be significantly smaller than disc electrode and disc electrode with Hjorth's method, especially at higher frequencies. This lower coherence in the high frequency band between neighboring tri polar electrodes suggests that the TCREs may record a more localized neuronal activity. The successful decoding of finger movements can provide extra degrees of freedom to drive brain computer interface (BCI) applications, especially for neurorehabilitation.
  • Thumbnail Image
    ItemOpen Access
    A multidisciplinary analytical approach to the identification of both modifiable and non-modifiable risk factors of dementia
    (Colorado State University. Libraries, 2021) Willoughby, Kathleen Angela, author; Davalos, Deana, advisor; Rojas, Don, committee member; Malinin, Laura, committee member; Cleveland, Jeanette, committee member
    In recent decades, dementia has become a growing global epidemic. As people are living longer, the number of individuals diagnosed with dementia has risen exponentially. Alzheimer's disease, the most common form of dementia, presently afflicts more that 5.4 million Americans (Thies et al., 2011). Though great strides have been made in dementia research, there is still much to be done to better pin-point disease risk and ameliorate decline and related symptom progression. This dissertation will focus on the efficacy of early intervention and risk factor identification as a first line of defense in staving off dementia progression. Within the B Sharp community-arts engagement program, we will evaluate domain-specific changes in older adult cognition over an acute and extended-duration timespan. Within the Alzheimer's Disease Neuroimaging Initiative, we will identify relevant risk factors associated with the consistent acceleration of cognitive decline as well as the slowing of such decline. As these proactive treatment approaches are more fully understood, better strategies for healthy aging can be implemented at both a generalized and individual level.
  • Thumbnail Image
    ItemOpen Access
    An investigation into the formation of representational associations in visual category learning
    (Colorado State University. Libraries, 2017) Jentink, Kade Garrett, author; Seger, Carol, advisor; Rojas, Don, committee member; Burzynska, Agnieszka, committee member
    Category learning allows us to use previous information we have accumulated, and extend it to new situations. Multiple systems are proposed to underlie learning, including: an explicit, rule-based system, and an implicit, procedural system. Information integration tasks are thought to load heavily onto the latter. In these tasks, a high degree of accuracy is reached only if participants can integrate incommensurable dimensions, often without being able to verbally describe how they are categorizing each stimulus. Learning in this type of task is thought to occur as participants associate a given stimulus with a category label, and then that label to a motor response. The present study sought to examine whether there may be an additional associative stage in which a stimulus is first associated with a "category representation" – a representation of the critical characteristics of a given category – which is then associated with a category label. Two experiments were conducted which attempted to determine whether this form of category representation is learned in information integration tasks. Both experiments reversed the category representation – category label association for a subset of stimuli and tested if subjects would transfer this reversal to the remaining stimuli, as should happen if they learned to associate each label with a single abstract category representation. Experiment 1 trained subjects with two sets of labels, each of which was associated with the same abstract category representation, to see if reversing one set of labels would alter the other. Experiment 2 trained subjects with 1 set of labels and tested if learning to reverse half of the stimulus space would transfer to the remaining half. In addition, the consistency of category label and motor response associations were manipulated in Experiment 2, with the hypothesis that subjects learning under inconsistent mappings would be forced to learn category labels and be more likely form an abstract category representation, whereas subjects learning under consistent conditions might only learn basic stimulus – response associations. Subjects in Experiment 1 did not transfer the reversal to the second set of category labels, inconsistent with the hypothesis that subjects would form an abstract category representation. However, over half the subjects in Experiment 2 did transfer reversed category label associations to untrained stimuli. Furthermore, a greater number of subjects transferred the reversals in the Inconsistent mapping condition. This is the first study to present evidence suggesting the existence of an abstract category representation and to provide a unique dissociation between consistent and inconsistent mappings for an information-integration task.
  • Thumbnail Image
    ItemOpen Access
    Cue competition and feature representation in a category learning task: an fMRI study
    (Colorado State University. Libraries, 2023) Jentink, Kade, author; Seger, Carol, advisor; Burzynska, Agnieszka, committee member; Rojas, Don, committee member; Thomas, Michael, committee member
    During learning, attention is limited, and therefore selecting what feature(s) to attend to in the environment is important. Sometimes, attention is captured by a cue or feature in such a way that other cues or features are not attended to, known as overshadowing. This process is not entirely understood in category learning, with some studies suggesting that it enhances learning of other features (Murphy et al., 2017), while others suggest that it inhibits (Lau et al., 2020). Furthermore, the location and organization of the neural representations that develop for category features during overshadowing has not been previously examined in this context. The present experiment used representational similarity analyses (RSA), a method for interrogating representational structure (Kriegeskorte et al., 2008), in order to examine where and how features were represented during overshadowing in a category learning task. Participants completed a category learning task in which categories were defined based on two informative features, one binary and one continuous. The binary feature was easier to learn (i.e., more salient), and it was hypothesized that it would overshadow learning of the more difficult continuous feature. This was demonstrated behaviorally: participants learned to categorize when the binary feature was present, then performed at chance when it was removed in a transfer task. Three different hypothetical models were fit to the neural data to determine underlying representational structure: a binary category model, an effector-specific motor model, and a model representing the degree of perceptual similarity for the continuous feature. During initial learning when the primary binary feature was present, the category model fit data from both early visual and object-specific areas of visual cortex, while the motor model fit data from motor-related regions including primary somatomotor cortex and the cerebellum. The perceptual similarity model for the continuous feature did not fit any task data during either Training or Transfer. However, there was a trend for the category model to fit activity in the basal ganglia and lateral occipital complex (LOC) during the Transfer task when the only information available for categorization was the continuous feature. Taken together, these results suggest that, although overshadowing inhibits use of the overshadowed continuous feature as the basis of categorization behavior, it might still contribute to activation of neural representations of category membership.
  • Thumbnail Image
    ItemOpen Access
    Large scale brain network mental workload engagement in schizophrenia
    (Colorado State University. Libraries, 2022) Duffy, John R., author; Thomas, Michael L., advisor; Rojas, Don, committee member; Blanchard, Nathanial, committee member; Tompkins, Sara Anne, committee member
    Objective: Cognitive deficits in patients diagnosed with schizophrenia are a core feature of the disorder. There are currently no treatments for these cognitive deficits. Our aim is to examine and compare patterns of increased versus decreased activity in the central executive network (CEN), salience network (SN), and default mode network (DMN) between healthy controls (HC) and patients diagnosed with schizophrenia (SZ) as well as to explore the influence of task load on these networks between HC and SZ. Method: Analyses focused on a secondary dataset comprising Blood Oxygen-Level Dependent (BOLD) data collected from 25 HC and 27 SZ who completed a working memory (WM) task (N-back) with 5 load conditions while undergoing functional magnetic resonance imaging (fMRI). Region of interest (ROI) data were analyzed using linear mixed-effects models. Dynamic causal modeling (DCM) was used in an exploratory analysis to examine working memory load input to these networks. Results: Group activation differences were found in the posterior salience network (pSN), default mode network (DMN), dorsal default mode network (dDMN), and ventral default mode network (vDMN) showing greater activity for SZ. Specifically, pSN, SMN, dDMN, and vDMN all showed increased activity in SZ compared to HC. The curve of brain activity was consistent between HC and SZ with the exception of the vDMN, where HC show greater activation at modest mental workload (quadratic curve) and SZ showed greater brain activation at lower mental workload (linear). In the CEN, there were no group differences, and the response curve was the same for both groups. In DCM analysis, working memory load acted as an input on different networks between HC and SZ. Conclusions: These group differences demonstrate network difference between HC and SZ and could show value in treatments targeting cognitive deficits in SZ from a large-scale brain network connectivity perspective. Future studies are needed to confirm these results with higher sample size in order to examine potential subtleties of interactions between these networks.
  • Thumbnail Image
    ItemOpen Access
    Mechanisms of timing: an integrative theoretical approach
    (Colorado State University. Libraries, 2019) Pantlin, Lara N., author; Davalos, Deana, advisor; Prince, Mark, advisor; Malcolm, Matthew, committee member; Rojas, Don, committee member
    Accurate timing allows individuals to perform essential tasks to meet societal demands, such as scheduling, responding to warning signals and planning. Since timing impacts various functions, understanding the meaning of a timing deficit is necessary. Poor performance in neurophysiological measures of timing has been related to psychopathology but has not specifically been related to one's ability to plan or maintain a schedule. Inability to track elapsed time as done in behavioral tasks is often related to poor performance in academic settings, but the intricacies of how inaccurate timing in one task manifests in other timing tasks has not been examined. The present study proposes a comprehensive examination of timing by dividing the field into three sub-domains: neurophysiological, behavioral, and applied temporal processing. These sub-domains are organized based on the tasks traditionally used to assess timing. Neurophysiological timing (Level I) was assessed using a duration-based mismatch negativity paradigm (dMMN), which fundamentally requires minimal cognitive resources. Behavioral timing (Level II) introduces the role of attention and working memory to accurately determine the amount of elapsed time (verbal estimation) or the generation of a pause, which reflects a specified amount of time (interval production). These tasks do not require the higher-order cognitive functions such as decision making and planning which are needed to accurately perform applied temporal processing tasks (e.g., time management and scheduling) (Level III). Hypothesis I proposed a hierarchical relationship among the three subdomains in which each succeeding level in the mediation is informed by the previous one and is distinct from the others based on the amount of cognition required to perform the task. Hypothesis II not only offered an extension of Hypothesis I, but also sought to examine the ways timing can be systematically improved through intervention methods. Across two time-points, participants were screened for select psychopathologies often associated with timing deficits (e.g., psychosis, traumatic brain injury, and substance use), underwent EEG recordings of dMMN to measure neurophysiology (Level I), performed two behavioral timing tasks (verbal estimation and interval production) (Level II), and completed three measures of applied temporal processing (letter-number sequencing and two time management surveys) (Level III). Hypothesis I was analyzed using a mediation model where neurophysiology (Level I) is expected to inform behavioral performance (Level II), which would subsequently influence accuracy on applied tasks (Level III). Hypothesis II was analyzed using repeated-measures ANOVAs to assess which intervention increases accuracy between time-points. Although Hypothesis I yielded nonsignificant results, interesting trends in the expected direction existed. Higher responses on the neurophysiological tasks were related to higher accuracy on behavioral and applied temporal processing measures. Hypothesis II yielded significant interactions between session and intervention and overall, suggested that using feedback to calibrate individuals to their abilities is the most appropriate intervention technique for increasing behavioral and applied accuracy. However, inclusion of tasks evaluating intermediate stages of timing is required if a full scale time continuum is to be modeled. Yet, this work provided the initial groundwork to further investigate the way time-related information is handled in the healthy brain.
  • Thumbnail Image
    ItemOpen Access
    Supervised and unsupervised training of deep autoencoder
    (Colorado State University. Libraries, 2017) Ghosh, Tomojit, author; Anderson, Charles, advisor; Kirby, Michael, committee member; Rojas, Don, committee member
    Deep learning has proven to be a very useful approach to learn complex data. Recent research in the fields of speech recognition, visual object recognition, natural language processing shows that deep generative models, which contain many layers of latent features, can learn complex data very efficiently. An autoencoder neural network with multiple layers can be used as a deep network to learn complex patterns in data. As training a multiple layer neural network is time consuming, a pre-training step has been employed to initialize the weights of a deep network to speed up the training process. In the pre-training step, each layer is trained individually and the output of each layer is wired to the input of the successive layers. After the pre-training, all the layers are stacked together to form the deep network, and then post training, also known as fine tuning, is done on the whole network to further improve the solution. The aforementioned way of training a deep network is known as stacked autoencoding and the deep neural network architecture is known as stack autoencoder. It is a very useful tool for classification as well as low dimensionality reduction. In this research we propose two new approaches to pre-train a deep autoencoder. We also propose a new supervised learning algorithm, called Centroid-encoding, which shows promising results in low dimensional embedding and classification. We use EEG data, gene expression data and MNIST hand written data to demonstrate the usefulness of our proposed methods.