AI-Assisted Treatment for Speech Sound Disorders

Overseen By Nina Benway, PhD

Age: < 18

Sex: Any

Trial Phase: Academic

Sponsor: Syracuse University

Must not be taking: Antiepileptics

Disqualifiers: Autism, Down Syndrome, ADHD, others

No Placebo Group All trial participants will receive the active study treatment (no placebo)

Compensation: Varies

Number of Visits: Unknown

Duration: 5 weeks

26 Participants Needed

Trial Summary

Will I have to stop taking my current medications?

The trial information does not specify whether participants need to stop taking their current medications. However, it does exclude those on antiepileptic medication, so it's best to discuss your specific medications with the trial coordinators.

What data supports the effectiveness of the AI-Assisted Treatment for Speech Sound Disorders?

Research shows that Speech Motor Chaining (SMC) can help children with speech sound disorders successfully learn new speech patterns and apply them to untrained words. Additionally, using technology like automatic speech recognition (ASR) for feedback has been effective in improving speech accuracy in individuals with speech disorders, suggesting that AI-assisted methods could enhance traditional speech therapy. ^{1 2 3 4 5}

How does the AI-Assisted Treatment for Speech Sound Disorders differ from other treatments?

The AI-Assisted Treatment for Speech Sound Disorders is unique because it uses artificial intelligence to provide real-time and objective feedback on speech performance, unlike traditional methods that rely on subjective assessments by speech-language pathologists. This approach can enhance the consistency and accuracy of speech therapy by analyzing speech patterns and providing tailored feedback to patients. ^{4 5 6 7 8}

What is the purpose of this trial?

The goal of this randomized-controlled trial is to determine how artificial intelligence-assisted home practice may enhance speech learning of the "r" sound in school-age children with residual speech sound disorders. All child participants will receive 1 speech lesson per week, via telepractice, for 5 weeks with a human speech-language clinician. Some participants will receive 3 speech sessions per week with an Artificial Intelligence (AI)-clinician during the same 5 weeks as the human clinician sessions (CONCURRENT treatment order group), whereas others will receive 3 speech sessions per week with an AI-clinician after the human clinician sessions end (SEQUENTIAL treatment order group.

Eligibility Criteria

This trial is for children aged 9 to 17 with speech sound disorders affecting their 'r' sounds, who speak American English and started learning it by age 3. They must have normal hearing, no significant cognitive or neurological conditions, no obstructive orthodontic appliances, and access to videoconferencing over broadband internet.

Inclusion Criteria

Must receive a percentile score of 8 or below on the Goldman-Fristoe Test of Articulation-3 (GFTA-3) Sounds in Words subtest

Must have reported hearing within normal limits

I want to improve how I pronounce the 'r' sound.

Exclusion Criteria

Must have no known history of autism spectrum disorder, Down Syndrome, cerebral palsy, intellectual disability, permanent hearing loss, epilepsy/antiepileptic medication, or brain injury/neurosurgery/stroke

Must have no orthodontic appliances that block the roof of the mouth (e.g., palate expanders)

Must not demonstrate childhood apraxia of speech (CAS-only) features in BOTH articulatory and rate/prosody domains of the ProCAD

Timeline

Screening

Participants are screened for eligibility to participate in the trial

2-4 weeks

Treatment

Participants receive 5 speech lessons with a human speech-language clinician, 1 time per week for 5 weeks. Concurrently or sequentially, they receive 15 speech lessons with an AI clinician, 3 times per week for 5 weeks.

5 weeks

Follow-up

Participants are monitored for changes in speech sound accuracy and retention after treatment

10 weeks

Treatment Details

Interventions

• Artificial Intelligence-led Speech Motor Chaining (CHAINING-AI)
• Speech-Language Pathologist-led Speech Motor Chaining

Trial Overview The study compares two methods of improving the 'r' sound in children: traditional lessons from a human speech-language pathologist once a week and additional sessions three times a week with an AI clinician either concurrently or sequentially.

Participant Groups

Group I: SEQUENTIAL treatment order Experimental Treatment 2 Interventions

* 5 speech lessons with a human speech-language clinician: 1 time per week for 5 weeks. * 15 speech lessons with an AI clinician (supervised by the caregiver), 3 times per week for the 5 weeks AFTER the human clinician sessions end.

Group II: CONCURRENT treatment order Experimental Treatment 2 Interventions

* 5 speech lessons with a human speech-language clinician: 1 time per week for 5 weeks. * 15 speech lessons with an AI clinician (supervised by the caregiver), 3 times per week DURING the same 5 weeks as the human clinician sessions.

Artificial Intelligence-led Speech Motor Chaining (CHAINING-AI) is already approved in United States for the following indications:

Approved in United States as ChainingAI for:

• Residual speech sound disorders, specifically for improving the 'r' sound in school-age children

Who Is Running the Clinical Trial?

Syracuse University

Lead Sponsor

Trials

54

Recruited

118,000+

National Institute on Deafness and Other Communication Disorders (NIDCD)

Collaborator

Trials

377

Recruited

190,000+

State University of New York - Upstate Medical University

Collaborator

Trials

176

Recruited

27,600+

National Institutes of Health (NIH)

Collaborator

Trials

2,896

Recruited

8,053,000+

Findings from Research

Ultrasound biofeedback therapy (UBT) shows promise in improving speech for children with residual speech sound disorder (RSSD), but it can be complex and demanding, necessitating simpler articulation targets for better understanding.

The study developed an image-analysis program, TonguePART, which successfully quantified tongue movements and distinguished accurate from misarticulated /ɑr/ sounds with up to 85% classification accuracy, highlighting the importance of tongue dorsum and blade movements for effective speech production.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Classification of accurate and misarticulated /&#593;r/ for ultrasound biofeedback using tongue part displacement trajectories. Li, SR., Dugan, S., Masterson, J., et al.[2023]

The study found that somatosensory inputs to oro-facial structures significantly improved speech processing for low-frequency words, indicating that these sensory cues can enhance lexical access and speech production accuracy.

In contrast, stimulation applied to non-speech areas (forehead) did not produce any significant effects, reinforcing the idea that targeted somatosensory interventions can effectively influence motor speech treatment outcomes.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Cross-Modal Somatosensory Repetition Priming and Speech Processing. Namasivayam, AK., Yan, T., Bali, R., et al.[2022]

The iPad-based speech therapy app using automatic speech recognition (ASR) showed an 80% agreement with human judgment on speech accuracy, indicating reliable feedback for users.

Participants with apraxia of speech and aphasia demonstrated significant improvements in word production accuracy over a 4-week therapy program, with gains maintained one month after treatment, suggesting the app's effectiveness as a therapeutic tool.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Feasibility of Automatic Speech Recognition for Providing Feedback During Tablet-Based Treatment for Apraxia of Speech Plus Aphasia. Ballard, KJ., Etter, NM., Shen, S., et al.[2020]

Speech Motor Chaining (SMC) is an effective treatment method for school-age children with speech sound disorders, promoting the acquisition of target speech patterns and generalization to untrained words.

Clinicians can implement SMC with over 90% fidelity, allowing children to complete more than 200 trials per session, which supports the method's practicality and potential for positive outcomes.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Tutorial: Speech Motor Chaining Treatment for School-Age Children With Speech Sound Disorders. Preston, JL., Leece, MC., Storto, J.[2020]

The multimodal speech capture system (MSCS) provides real-time and offline visualization of tongue and lip movements along with voice, offering immediate feedback to patients and allowing speech-language pathologists (SLPs) to analyze articulation more effectively.

This innovative system aims to enhance speech rehabilitation by delivering objective, quantitative feedback on speech performance, which could lead to more consistent and effective treatment compared to traditional subjective methods.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Multimodal Speech Capture System for Speech Rehabilitation and Learning. Sebkhi, N., Desai, D., Islam, M., et al.[2018]

Machine learning techniques, particularly Light Gradient Boosting, can accurately detect Parkinson's Disease using just voice recordings, achieving a high accuracy rate (AUC of 0.951) with only seven acoustic features.

This study analyzed data from 80 subjects, including 40 patients with Parkinson's Disease and 40 controls, demonstrating that non-invasive voice analysis could serve as an effective screening tool for early detection of the disease.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Gradient boosting for Parkinson's disease diagnosis from voice recordings. Karabayir, I., Goldman, SM., Pappu, S., et al.[2021]

Machine learning algorithms (MLAs) can accurately classify patients with Parkinson's disease (PD) and non-PD patients based on voice features, achieving over 90% accuracy.

MLAs can enhance speech rehabilitation for PD patients by converting unclear speech into intelligible speech and automatically assessing speech features, potentially offering a more effective treatment option than traditional therapies.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Management of Parkinson's Disease Dysarthria: Can Artificial Intelligence Provide the Solution? Kumar, R., Tripathy, M., Kumar, N., et al.[2022]

Advancements in computerized automatic speech recognition (ASR) technology have made it more reliable and user-friendly, expanding its applications beyond simple transcription to various fields, including voice and speech therapy.

ASR technology is expected to transform human-computer interaction by replacing traditional input devices like keyboards and mice with voice commands, allowing for more accessible communication for individuals with oral communication disorders.

NCBI (Pubmed)

pubmed.ncbi.nlm.nih.gov

Automatic speech recognition (ASR) and its use as a tool for assessment or therapy of voice, speech, and language disorders. Kitzing, P., Maier, A., Ahlander, VL.[2018]

References

1. England pubmed.ncbi.nlm.nih.gov

Classification of accurate and misarticulated /&#593;r/ for ultrasound biofeedback using tongue part displacement trajectories. [2023]

2. Singapore pubmed.ncbi.nlm.nih.gov

Cross-Modal Somatosensory Repetition Priming and Speech Processing. [2022]

3. United States pubmed.ncbi.nlm.nih.gov

Feasibility of Automatic Speech Recognition for Providing Feedback During Tablet-Based Treatment for Apraxia of Speech Plus Aphasia. [2020]

4. United States pubmed.ncbi.nlm.nih.gov

Tutorial: Speech Motor Chaining Treatment for School-Age Children With Speech Sound Disorders. [2020]

5. United States pubmed.ncbi.nlm.nih.gov

Multimodal Speech Capture System for Speech Rehabilitation and Learning. [2018]

6. England pubmed.ncbi.nlm.nih.gov

Gradient boosting for Parkinson's disease diagnosis from voice recordings. [2021]

7. India pubmed.ncbi.nlm.nih.gov

Management of Parkinson's Disease Dysarthria: Can Artificial Intelligence Provide the Solution? [2022]

8. England pubmed.ncbi.nlm.nih.gov

Automatic speech recognition (ASR) and its use as a tool for assessment or therapy of voice, speech, and language disorders. [2018]