Objective:

To integrate evidence on eye-movement-based and multimodal computational methods for dyslexia screening and risk identification during reading tasks.

Approach:

Key Findings:

• Twenty-three articles included from 50 screened, covering various study types related to eye-movement biomarkers and machine-learning predictions.
• Dyslexic readers showed longer fixation durations, increased regression behavior, and reduced saccadic efficiency.
• Machine-learning algorithms achieved classification accuracies ranging from 80% to 95%, with some studies reporting accuracies approaching 99% under specific conditions.

Interpretation:

There is significant heterogeneity in datasets, feature extraction methods, and validation schemes, with many studies using proxy diagnostic labels and small datasets.

Limitations:

• Heterogeneity in datasets and methods.
• Many studies relied on small or internally derived datasets.
• Limited multimodal evidence, with most studies focusing on eye-movement data.

Conclusion:

Eye-movement based computational systems are a promising, non-invasive method for scalable dyslexia screening.

Sources:

• Systematic Review Registration
Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.