Using a computer model to guide treatment for acquired dyslexia

Evidence-based Modeling for Acquired Dyslexia Treatments

This trial will test whether a computer (neural network) model can predict which of two 60-hour therapies will better help people with new reading problems after a left-hemisphere stroke.

Quick facts

• Phase: NA
• Study type: Interventional
• Enrollment: 12 (estimated)
• Ages: 18 Years to 85 Years
• Sex: All
• Sponsor: Rutgers, The State University of New Jersey (other)
• Locations: 3 sites (Tallahassee, Florida and 2 other locations)
• Trial ID: NCT07209488 on ClinicalTrials.gov

What this trial studies

This is a low-risk, early phase 1, multicenter trial that uses a computational neural-network model of reading to simulate acquired dyslexia and to predict which of two therapies should work best for each person. All participants receive two full rounds of treatment — each round is either phonomotor therapy (PMT) or semantic feature analysis (SFA), delivered as 60 hours of therapy spread across 2 hours a day, 5 days a week. The trial compares outcomes when participants receive the treatment the model predicts as advantageous versus the alternative treatment. Reading ability and related language measures will be used to determine whether model-guided treatment produces greater improvement.

Who should consider this trial

Good fit: Adults who were native English speakers before their left-hemisphere stroke, have normal or corrected vision and hearing, left-hemisphere ischemic or hemorrhagic stroke confirmed by brain scan, and measurable reading impairment on the Woodcock Reading Mastery Test - III Basic Skills cluster are ideal candidates.

Not a fit: People with pre-stroke developmental dyslexia, other pre-stroke neurological diseases, severe apraxia of speech, major current psychiatric disorders, or who cannot provide or undergo post-stroke brain imaging may not benefit or may be excluded.

Why it matters

Potential benefit: If successful, the approach could help personalize therapy to improve reading recovery after stroke by identifying which therapy is most likely to help each person.

How similar studies have performed: Behavioral therapies like phonomotor treatment and semantic feature analysis have shown benefit for post-stroke language and reading problems, but using a neural-network model to predict which therapy to give is a novel approach that has not been widely tested in clinical trials.

Eligibility criteria

Inclusion Criteria:

* Fluent in English as a first language pre-stroke by self report.
* Normal or corrected-to-normal vision.
* Normal or aided hearing.
* Left hemisphere ischemic or hemorrhagic stroke (as verified by a structural brain scan).
* Impaired reading as confirmed by significant impairment in the Basic Skills cluster of the Woodcock Reading Mastery Test - III.

Exclusion Criteria:

* Diagnosed pre-stroke neurological disease affecting the brain other than left hemisphere stroke.
* Severe apraxia of speech (determined by consensus judgment among speech-language pathologists).
* History of learning disabilities such as developmental dyslexia or current self-reported major psychiatric disorders.
* Inability to undergo, or provide a copy of, a post-stroke brain imaging scan.

Where this trial is running

Tallahassee, Florida and 2 other locations

• Florida State University — Tallahassee, Florida, United States (RECRUITING)
• Johns Hopkins University School of Medicine — Baltimore, Maryland, United States (RECRUITING)
• Kessler Foundation — West Orange, New Jersey, United States (RECRUITING)

Study contacts

• Principal investigator: William W Graves, PhD — Rutgers, The State University of New Jersey
• Study coordinator: William W Graves, PhD

Conditions: Stroke, Dyslexia, Acquired, reading, computational model, treatment, therapy