Electroencephalography (EEG) provides valuable insight into the neural mechanisms underlying dyslexia, yet analysis is challenged by low signal to noise ratio (SNR), high inter-subject variability, and complex spatio-temporal dynamics. The Neural Congruency framework offers a promising way to identify consistent brain activity patterns among readers, but its integration with deep learning remains limited. This study introduces Neural Congruency Contrastive Learning (NCCL), a framework that combines spatial, frequency, and temporal convolutions to learn EEG embeddings aligned with neural congruency principles. Using synthetic EEG representing dyslexic and control participants across SNRs from -37 dB to -7 dB, the model was trained with a contrastive loss to maximize within group similarity and enhance between group separation. NCCL reliably distinguished dyslexic from control groups even at -25 dB, showing high stability across runs and maintaining discriminative performance under severe noise. These results highlight the framework’s robustness and its potential applicability to real EEG datasets, including tasks such as Rapid Automatized Naming and Phonological Awareness. Overall, this work establishes a noise resilient approach for modeling neural congruency with deep contrastive learning, advancing the use of artificial intelligence in dyslexia research and future clinical assessment.

Understanding the relationship between neural activations and cognitive factors in reading disorders, such as dyslexia, is a key goal in developmental neuroscience. Despite advancements in computational methods for extracting neural components at the aggregate, inter-subject variability in EEG signals make it challenging to perform across-group single-trial analysis on the neural mechanisms associated with reading disorders. Our contribution focuses on using the Neural-Congruency framework to address the issue of inter-subject variability in EEG data, which allows for single-trial analysis across groups of dyslexic and control participants. Preliminary results demonstrate the feasibility of this approach, highlighting its potential to deepen our understanding of the neural mechanisms underlying reading disorders. By providing a more granular and accurate analysis of neural activations, this study lays the groundwork for improved diagnostic and intervention strategies for dyslexia.

The study’s primary contribution is a game-based cognitive intervention tool for young learners with reading difficulties. The second contribution is the game designer's guide tool, which provides a set of recommended design guidelines for designers, developers and researchers that create games for children. The tools were designed and developed in the ReaDi-STANCE project and are accessible on the project’s platform (ReaDi-STANCE platform home page: https://www.cs.ucy.ac.cy/projects/readistance/ ). The platform has the following main modules integrated: a management support module, an algorithm module to rotate in a predefined order the games the child end-user plays, a prompting stages module to support the child end-user during gameplay, and a data analytics module that researchers can use. Two different groups of experts conducted two evaluations to evaluate: (1) a corpus of game design themes and guidelines for children, extracted from the respective literature, to develop the game designer's guide tool and (2) the extent of application of the design above guidelines in the game-based cognitive intervention tool. We present the intervention impact results and discuss their effect on improving reading skills in poor readers.

Electroencephalogram (EEG) recordings of children are often used to study the underlying neural basis of causal factors of reading disorders and dyslexia. However, the inter-subject variability in EEG and the unconstrained nature of reading experiments used to elicit these factors made it challenging for traditional EEG analysis methods to extract neural components of these factors. In this work, we aim to explore the use of novel deep neural network architectures and contrastive learning methods to overcome the methodological limitations of traditional techniques and enhance the extraction process of neural components during reading tasks. Notably, we formulate a neural network architecture to extract EEG embedding using contrastive loss that maximizes the neural congruency in non-dyslexic children compared to children with dyslexia. We plan to evaluate our approach on three EEG datasets involving children with dyslexia performing Rapid Automatized Naming (RAN) and Phonological Processing (PA) tasks. The proposed contrastive learning framework will provide an enhanced tool to facilitate studying the neural underpinnings of naming speed and their association with reading performance and related difficulties.

The present study followed a reading-level match design to investigate group differences in eye movements between grade three and grade six Greek-speaking children with reading difficulties (RD) and controls (chronological age (CA) and reading-level (RL)-matched groups), examining their performance on RAN tasks of different modalities (phonological vs. visual) and complexity levels (confounding vs. not-confounding conditions). Three eye movements (fixations, saccades, and regressions) were recorded using the EyeLink 1000 Plus eye-tracking system. The results showed that both grade three and grade six RD groups produced more and longer durations and regressions and more saccades compared to their CA controls in all tasks. However, no differences were observed between the grade six RD and the RL-matched groups in the eye-tracking measures. The present findings have important implications for determining the contribution of the reading level match design in eye-tracking reading-related research and exploring the causality of reading difficulties in consistent orthographies.

Measuring simultaneous processing, a reliable predictor of reading development and reading difficulties (RDs), has traditionally involved cognitive tasks that test reaction or response time, which only capture the efficiency at the output processing stage and neglect the internal stages of information processing. However, with eye-tracking methodology, we can reveal the underlying temporal and spatial processes involved in simultaneous processing and investigate whether these processes are equivalent across chronological or reading age groups. This study used eye-tracking to investigate the simultaneous processing abilities of 15 Grade 6 and 15 Grade 3 children with RDs and their chronological-age controls (15 in each Grade). The Grade 3 typical readers were used as reading-level (RL) controls for the Grade 6 RD group. Participants were required to listen to a question and then point to a picture among four competing illustrations demonstrating the spatial relationship raised in the question. Two eye movements (fixations and saccades) were recorded using the EyeLink 1000 Plus eye-tracking system. The results showed that the Grade 3 RD group produced more and longer fixations than their CA controls, indicating that the pattern of eye movements of young children with RD is typically deficient compared to that of their typically developing counterparts when processing verbal and spatial stimuli simultaneously. However, no differences were observed between the Grade 6 groups in eye movement measures. Notably, the Grade 6 RD group outperformed the RL-matched Grade 3 group, yielding significantly fewer and shorter fixations. The discussion centers on the role of the eye-tracking method as a reliable means of deciphering the simultaneous cognitive processing involved in learning.

Objective: We explored neural components in Electroencephalography (EEG) signals during a phonological processing task to assess (a) the neural origins of Baddeley's working-memory components contributing to phonological processing, (b) the unitary structure of phonological processing and (c) the neural differences between children with dyslexia (DYS) and controls (CAC). Methods: EEG data were collected from sixty children (half with dyslexia) while performing the initialand final- phoneme elision task. We explored a novel machine-learning-based approach to identify the neural components in EEG elicited in response to the two conditions and capture differences between DYS and CAC. Results: Our method identifies two sets of phoneme-related neural congruency components capturing neural activations distinguishing DYS and CAC across conditions. Conclusions: Neural congruency components capture the underlying neural mechanisms that drive the relationship between phonological deficits and dyslexia and provide insights into the phonological loop and visual-sketchpad dimensions in Baddeley's model at the neural level. They also confirm the unitary structure of phonological awareness with EEG data. Significance: Our findings provide novel insights into the neural origins of the phonological processing differences in children with dyslexia, the unitary structure of phonological awareness, and further verify Baddeley's model as a theoretical framework for phonological processing and dyslexia. (c) 2023 International Federation of Clinical Neurophysiology. Published by Elsevier B.V. All rights reserved.

Reading intervention program efficacy is usually determined by comparing participants' performance to controls on dependent measures at pre-, mid-, and post-intervention assessments. However, little is known about how learning progresses during different stages of the intervention. This lack of knowledge can be attributed to the absence of appropriate computational frameworks to encode, analyze, and capture such dynamics. We propose a novel computational framework to capture learning process dynamics during the intervention by analyzing microgenetic data. The framework addresses the problem of encoding microgenetic data into a common data representation model, introduces four information-theoretic metrics to capture the instantaneous developmental learning stages of groups and individuals, and provides the mathematical model to analyze those metrics for the study of learning stages during the intervention. We used data from a longitudinal reading remediation study involving 56 Greek-speaking 6-year-old children to demonstrate the framework's utility. Results showed that the framework functions as a new tool to explore the modulation in learning stages during the intervention, better understand how reading occurs, and how reading disability may be adequately treated.

Children's performance on the spoonerism task, a behavioral test that measures phonological processing skills, predicts reading abilities and related disorders.However, this relationship between phonological processing skills and dyslexia has been primarily examined based on behavioral responses to the spoonerism task.As a result, there is a growing interest in developmental neuroscience to explore the neural origins of this relationship and its relation to reading difficulties.Yet, traditional electroencephalography (EEG) analysis methods had little success identifying informative neural components that depict neural differences in children with reading disorders during spoonerism.The current study explores a novel computational approach to isolate informative neural signatures elicited during the spoonerism test.We apply our method to EEG data obtained from a group of children with dyslexia and controls during the execution of a spoonerism task.Our findings demonstrate that our method extracts components that characterize the neural origins of complex cognitive phonological processes, explains differences between children with dyslexia and controls, and generates novel insights into the neural underpinnings of dyslexia in children.

ObjectiveNaming speed, behaviorally measured via the serial Rapid automatized naming (RAN) test, is one of the most examined underlying cognitive factors of reading development and reading difficulties (RD). However, the unconstrained-reading format of serial RAN has made it challenging for traditional EEG analysis methods to extract neural components for studying the neural underpinnings of naming speed. The present study aims to explore a novel approach to isolate neural components during the serial RAN task that are (a) informative of group differences between children with dyslexia (DYS) and chronological age controls (CAC), (b) improve the power of analysis, and (c) are suitable for deciphering the neural underpinnings of naming speed. MethodsWe propose a novel machine-learning-based algorithm that extracts spatiotemporal neural components during serial RAN, termed RAN-related neural-congruency components. We demonstrate our approach on EEG and eye-tracking recordings from 60 children (30 DYS and 30 CAC), under phonologically or visually similar, and dissimilar control tasks. ResultsResults reveal significant differences in the RAN-related neural-congruency components between DYS and CAC groups in all four conditions. ConclusionRapid automatized naming-related neural-congruency components capture the neural activity of cognitive processes associated with naming speed and are informative of group differences between children with dyslexia and typically developing children. SignificanceWe propose the resulting RAN-related neural-components as a methodological framework to facilitate studying the neural underpinnings of naming speed and their association with reading performance and related difficulties.

The present study followed a reading-level match design to investigate group differences in eye movements between grade three and grade six Greek-speaking children with reading difficulties (RD) and controls (chronological age (CA) and reading-level (RL)-matched groups), examining their performance on RAN tasks of different modalities (phonological vs. visual) and complexity levels (confounding vs. not-confounding conditions). Three eye movements (fixations, saccades, and regressions) were recorded using the EyeLink 1000 Plus eye-tracking system. The results showed that both grade three and grade six RD groups produced more and longer durations and regressions and more saccades compared to their CA controls in all tasks. However, no differences were observed between the grade six RD and the RL-matched groups in the eye-tracking measures. The present findings have important implications for determining the contribution of the reading level match design in eye-tracking reading-related research and exploring the causality of reading difficulties in consistent orthographies.

Identifying the neural basis of dyslexia is a fundamental goal of developmental neuroscience. Final-phoneme elision (PE) test is a paradigm used for assessing phonological deficit (PD), which is widely considered a causal risk factor for dyslexia. However, the causal relationship between PD to dyslexia has been examined primarily based on behavioral observations. Towards facilitating the exploration of the neurophysiological origins of the theorized link between PD and dyslexia, we set out to isolate differential neural activation patterns in children with dyslexia during PE. Accordingly, we present a machine-learning-based approach to identifying differential brain activity in childrenwith dyslexia and controls during the PE. Our method formulates an optimization problem to extract informative EEG components based on the `Neural-congruency hypothesis', termed Phoneme-related Neural-congruency components. It then uses amachine-learning algorithm to optimally combine the resulting components to differentiate between the neural activity of children with dyslexia and controls. We apply our approach to a real EEG dataset involving children with dyslexia and controls. Our findings demonstrate that our method generates novel insights into the neural underpinnings of dyslexia and the potential neural origins of phonological deficits as a causal factor of dyslexia. Notably, our approach overcomes several methodological challenges in conventional EEG analysis methods; therefore, it could be utilized in studying the neural origins of other behaviorally defined developmental disorders previously overlooked because of such methodological constraints.

Understanding the neural underpinnings of dyslexia is an open and fundamental question in developmental neuroscience. A widely agreed causal risk factor for dyslexia is phonological deficit (PD). However, the causal relationships between PD and dyslexia have been primarily investigated and theorized based on findings derived from behavioral measures. What is missing is evidence of the underlying neurophysiological origins of these relationships. The present study examined whether the performance on a phonological awareness task, namely phoneme elision (PE), differentiated children with dyslexia from their typically developing counterparts at a neural level. We proposed a novel machine-learning-based approach to extract neural activity from EEG to identify neural differences at the group level. Specifically, we formulated an optimization problem to first extract informative EEG components (termed phoneme-related neural-congruency components) by maximizing the congruency in neural activity among typically developing children during phoneme elision. Next, we utilized a machine-learning algorithm to optimally combine the resulting components to differentiate between children with dyslexia and controls. Results showed that the proposed phoneme-related neural-congruency components are predictive about the underlying neuronal differences amongst groups. These results provide empirical evidence towards the neural underpinnings of dyslexia and the potential neural origins of PD as a causal link to dyslexia. Notably, the proposed method could be used to study other behaviorally defined developmental disorders.

Emotions affect our decisions, experiences, preferences, and perceptions. Understanding the neural underpinning of human emotions is a fundamental goal of neuroscience research. Moreover, EEG-based emotion recognition is a key component towards the development of affective-aware intelligent systems. However, characterizing the neural basis of emotions elicited during video viewing has been proven a challenging task. In this paper, we propose a novel machine-learning approach to isolate neural components in EEG signals that are informative of the affective content of emotionally-loaded videos. Based on these components, we define a set of neural metrics and evaluate them as potential indicators of the overall emotional content of each video. We demonstrate the predictive power of the proposed metrics, on the DEAP benchmark dataset for EEG-based emotion recognition. Our results provide novel empirical evidence that the neural components extracted by our method can serve as an informative metric in EEG-based emotion recognition during video viewing and achieving a 4-fold increase in predictive power compared to traditional frequency-based metrics. Moreover, each extracted component is associated with a spatial and a temporal profile, that allows researchers to inspect and interpret the spatiotemporal origins of the underlying neural activity. Thus, our method a framework that facilitates the study of neural correlates of emotion during video viewing.

This paper focuses on the application of computer vision and convolutional neural network techniques in the automotive industry to reduce the amount of time required to locate a vacant parking spot and to reduce driving time. The main motivation for a vacant parking spot detector is such that today’s drivers are facing major difficulties in finding available spots in largely populated cities. This often time leads to increased congestion and frustration for the driver because they are forced to continue their search for a parking spot. Our approach is able to solve this issue and provide the driver with useful information through the use of transfer learning methodologies. The main contribution of this paper is to examine and improve on previously implemented transfer learning methods in order to better increase the detection accuracy. This paper differs from previous attempts such that it considers all environmental factors such as weather and time of day. Other models are not able to handle these conditions with a high accuracy and subsequently falter. When compared to previous attempts, our implementation focuses solely on the reliance of transfer learning. The results indicate that our model is capable of identifying vacant parking spaces under all conditions with competitive accuracies. The proposed model is able to surpass the accuracy of the latest attempt at solving this issue.

To inform a proper diagnosis and understanding of Alzheimer’s Disease (AD), deep learning has emerged as an alternate approach for detecting physical brain changes within magnetic resonance imaging (MRI). The advancement of deep learning within biomedical imaging, particularly in MRI scans, has proven to be an efficient resource for abnormality detection while utilizing convolutional neural networks (CNN) to perform feature mapping within multilayer perceptrons. In this study, we aim to test the feasibility of using three-dimensional convolutional neural networks to identify neurophysiological degeneration in the entire-brain scans that differentiate between AD patients and controls. In particular, we propose and train a 3D-CNN model to classify between MRI scans of cognitively-healthy individuals and AD patients. We validate our proposed model on a large dataset composed of more than seven hundred MRI scans (half AD). Our results show a validation accuracy of 79% which is at par with the current state-of-the-art. The benefits of our proposed 3D network are that it can assist in the exploration and detection of AD by mapping the complex heterogeneity of the brain, particularly in the limbic system and temporal lobe. The goal of this research is to measure the efficacy and predictability of 3D convolutional networks in detecting the progression of neurodegeneration within MRI brain scans of HC and AD patients.

OBJECTIVE:We combined electroencephalography (EEG) and eye-tracking recordings to examine the underlying factors elicited during the serial Rapid-Automatized Naming (RAN) task that may differentiate between children with dyslexia (DYS) and chronological age controls (CAC).METHODS:Thirty children with DYS and 30 CAC (Mage = 9.79 years; age range 7.6 through 12.1 years) performed a set of serial RAN tasks. We extracted fixation-related potentials (FRPs) under phonologically similar (rime-confound) or visually similar (resembling lowercase letters) and dissimilar (non-confounding and discrete uppercase letters, respectively) control tasks.RESULTS:Results revealed significant differences in FRP amplitudes between DYS and CAC groups under the phonologically similar and phonologically non-confounding conditions. No differences were observed in the case of the visual conditions. Moreover, regression analysis showed that the average amplitude of the extracted components significantly predicted RAN performance.CONCLUSION:FRPs capture neural components during the serial RAN task informative of differences between DYS and CAC and establish a relationship between neurocognitive processes during serial RAN and dyslexia.SIGNIFICANCE:We suggest our approach as a methodological model for the concurrent analysis of neurophysiological and eye-gaze data to decipher the role of RAN in reading.

In the current age of coronavirus, monitoring and enforcing correct mask-wearing regulation in public spaces is of para- mount importance. Specifically, there is a need to monitor whether people wear masks and whether they wear them cor- rectly. However, there is a lack of automated systems to rec- ognize correct mask-wearing compliance. In this paper, we propose a computer-vision-based solution to the problem of mask-wearing monitoring. In particular, we propose a convo- lutional neural network to recognize images of people wear- ing masks correctly, people wearing masks incorrectly, and people not wearing masks at all. Our proposed model is shown to predict correct mask-wearing practices with over 98% accuracy. The model can be easily deployed as an auto- mated system to screen people entering indoor spaces, and can replace current manual, time-consuming, temperature- screening practices. Such applications can serve as an im- portant tool to help reduce transmission rates during the cur- rent pandemic.

Understanding the neural underpinning of reading disorders, such as dyslexia, is a fundamental question in developmental neuroscience. However, identifying and isolating informative neural components elicited during free-naming paradigms (i.e. unprompted and unconstrained naming tasks) has proven a challenging methodological task. These methodological barriers have hindered the study of the neural underpinnings of reading disorders. In this paper, we proposed a machine learning approach for detecting neural components during free-naming, overcoming much of the current methodological challenges. We propose a new neural-based metric to differentiate groups of children with dyslexia (DYS) and their chronological age controls (CAC) in a free-naming task. Our approach combines electroencephalography (EEG) and eye-tracking measures to generate single-trial fixation-related potentials (sFRPs) and formulate an optimization problem to extract naming-related neural components, informative of group differences. Our approach is validated on a real dataset involving children with dyslexia and CAC performing a Rapid-Automatized Naming (RAN) task. Our results demonstrate the validity of the proposed metric as an indicator of the neural-based markers of reading disorders. Importantly, our proposed framework provides a novel approach that can facilitate the study of neural correlates of reading disorders under paradigms current methods are unable to.

Emotions affect our decisions, experiences, preferences, and perceptions. Understanding the neural underpinning of human emotions is a fundamental goal of neuroscience research. Moreover, EEG-based emotion recognition is a key component towards the development of affective-aware intelligent systems. However, characterizing the neural basis of emotions elicited during video viewing has been proven a challenging task. In this paper, we propose a novel machine-learning approach to isolate neural components in EEG signals that are informative of the affective content of emotionally-loaded videos. Based on these components, we define a set of neural metrics and evaluate them as potential indicators of the overall emotional content of each video. We demonstrate the predictive power of the proposed metrics, on the DEAP benchmark dataset for EEG-based emotion recognition. Our results provide novel empirical evidence that the neural components extracted by our method can serve as an informative metric in EEG-based emotion recognition during video viewing and achieving a 4-fold increase in predictive power compared to traditional frequency-based metrics. Moreover, each extracted component is associated with a spatial and a temporal profile, that allows researchers to inspect and interpret the spatiotemporal origins of the underlying neural activity. Thus, our method a framework that facilitates the study of neural correlates of emotion during video viewing.