Brain Connectivity Analysis Based on Reading Tasks in Children with Dyslexia (Using EEG Signals)

Dyslexia is a neurodevelopmental disorder that affects the normal reading ability, even though children with normal intelligence may still be affected. This paper investigates the differences in brain connectivity between children with dyslexia and normal children during reading tasks, and analyzes these differences using graph theory. The study examines the brain functional connectivity of children with dyslexia and control groups during reading tasks, providing possible evidence of brain network impairment.

Background

Developmental Dyslexia (DD) is a neurodevelopmental reading disorder that affects approximately 5% to 10% of the population. Despite having normal intelligence, these children show significant gaps in academic performance. Understanding the neurophysiological causes of dyslexia and early detection is crucial to prevent academic challenges and psychological issues for these children. Although there are numerous behavioral studies showing differences in reading skills between individuals with dyslexia and normal individuals, the root cause remains unclear, necessitating further research to uncover the neurophysiological differences in the brains of those with dyslexia.

With the advancement of neuroimaging methods, researchers can better understand the brain functions related to various neurological disorders. Different neuroimaging techniques have been used to study the brain activity differences between individuals with dyslexia and normal individuals under resting and task states. Functional imaging studies consistently report less or reduced activity in the left hemisphere parietal-temporal regions of children with dyslexia. However, recent literature suggests that disruptions in functional connectivity between reading areas may be the cause of DD.

Paper Information

This study, authored by Guhan Seshadri N.P. and Bikesh Kumar Singh from the Biomedical Engineering Department at the National Institute of Technology Raipur, India, was accepted on March 21, 2024, and published in the same year in the journal Medical & Biological Engineering & Computing, which is published by the International Federation for Medical and Biological Engineering.

Research Methods

Data Collection and Task Details

The study involved 15 children with dyslexia and 15 normally developing children, with the former coming from special schools and the latter from regular elementary schools. All children were assessed by psychologists to ensure an IQ above 85 and no history of hearing or neurological diseases.

A 19-channel EEG system was used, with electrodes placed on the scalp according to the 10-20 system, a sampling rate of 256Hz, and a recording bandwidth between 0.1 to 70Hz. Each child performed a reading task involving two different trials, each containing two stimuli: one word presented as an image (e.g., “sand”), and the other as an audio of the same or different word (e.g., “sand” or “land”). Children responded by pressing buttons to match the stimuli.

Data Processing and Analysis

Firstly, the collected EEG signals underwent preprocessing, including applying a three-point moving average filter and wavelet denoising techniques to reduce noise. The EEG signals were decomposed into frequency bands, extracting different frequency bands such as δ (0-4Hz), θ (4-8Hz), α (8-13Hz), β (13-32Hz), and γ (32-64Hz).

Graph Network Analysis

Brain networks were studied using graph theory. A 19×19 functional connectivity matrix was generated for each subject based on EEG signal coherence, and network features were extracted, including node strength, characteristic path length (PL), clustering coefficient (CL), global efficiency (EG), local efficiency (EL), and small-world network (SW).

Statistical Analysis

Non-parametric Mann-Whitney U tests were used to statistically compare the graph features between the two groups, and Wilcoxon signed-rank tests were used to analyze the significance within-group conditions. The results indicated significant differences in PL, CL, EG, and EL in θ and α bands between the dyslexic group and the control group.

Results

Performance on Reading Tasks

The control group’s performance accuracy on the tasks was significantly better than the dyslexic group, with the dyslexic group having longer response times (RT).

Network Measurement Results

Network Measurements in the Dyslexic Group

In the task state, the δ band strength at t5 and t3 electrode positions in the dyslexic group was significantly higher, while the strength in θ and α bands was significantly lower. Similarly, in terms of PL, CL, EG, and EL, the dyslexic group was significantly higher than the baseline values during the task period, reflecting deficiencies in information transmission and local processing networks.

Network Measurements in the Control Group

The control group had significantly higher strengths in the θ, α, and β bands during the task state, with significantly lower PL and significantly higher EG and EL, indicating more efficient information transmission and processing capabilities during the task.

Comparison Between the Two Groups

Compared to the control group, the dyslexic group had significantly higher δ band strength and significantly lower θ, α, and β band strengths in the task state. PL was longer, CL and EG were lower, indicating that the brain network of the dyslexic group was impaired during reading tasks, with lower integration and segregation reflecting reduced functional efficiency.

Regional Network Activation Differences

During the task, δ band strength at t5 and t4 electrode positions was significantly higher in the dyslexic group, while θ, α, and β band strengths were lower, reflecting deficiencies in information encoding, processing, and working memory.

Discussion

This study used graph theory to investigate the brain functional connectivity of dyslexic and normal children during reading tasks. The findings revealed that the brain network organization of the dyslexic group was disrupted during the task period, displaying lower functional integration and segregation. The results demonstrate the effectiveness of graph theory in understanding the brain network and neurophysiological basis of dyslexic children.

Conclusion

This study shows that graph theory analysis based on EEG signal functional connectivity can reveal brain network impairments in children with dyslexia during reading tasks and provides an effective way to understand their neurophysiological basis. This finding may provide new references for early detection and intervention in dyslexia.

Through detailed experimental design and data analysis methods, this study provides new insights into the neurophysiological basis of children with dyslexia and demonstrates the potential application of graph theory in brain network research.