In Fall 2023, Vanderbilt Data Science hosted an AI Deep Dive session with Dr. Matthew Pontell, focusing on generative AI’s potential in enhancing speech therapy and improving speech dysfunction outcomes. These sessions are an excellent opportunity to delve into research problems that could be addressed with deep learning, engaging with data scientists, researchers, and students from across the university to brainstorm and formulate solutions.
Bridge2AI Voice Conference Presentation
Our students recently presented their findings at the Bridge2AI Voice conference hosted by the NIH in Tampa, FL. The poster titled “Leveraging Large Language Models for Hypernasality Classification in Cleft Palate Patients” showcased the project’s progress. The event was a blend of professionals from various fields, discussing voice technologies for patient care. Our team not only presented their work but also attended workshops and enjoyed activities like kayaking in Tampa Bay.
Session Highlights:
On October 27th, Dr. Pontell led a discussion on using generative AI to improve speech health. The session shed light on innovative strategies and recovery techniques that promise better outcomes for those suffering from speech impairments.
- Purpose: To explore how generative AI can aid in treating speech dysfunctions and create individualized therapy plans.
- Innovative Techniques: Dr. Pontell shared advancements in generative AI used to simulate and analyze speech patterns, providing targeted feedback for therapy.
- AI Applications: Discussions included AI technology’s role in supporting therapists and enabling patients to practice and improve speech independently.
Session Insights:
- Generative AI has the potential to transform speech therapy by providing personalized, adaptable tools for both diagnosis and treatment.
- AI applications in speech health can significantly accelerate the therapy process, offering real-time analysis and feedback.
Project Update: The Cleft Palate Project
The Cleft Palate Project is developing AI-driven user interfaces that simulate and analyze speech patterns to provide targeted feedback for therapy. This project, inspired by Dr. Pontell’s Deep Dive session, aims to train an audio model to identify speech impediments and create individualized therapy plans. Key features include:
- User Interface: Accessible via a web app for testing and deployment.
- Phone Interface: Usable worldwide through a phone call line for voice sample recording and analysis.
Principal Investigators:
- Dr. Matthew Pontell, Assistant Professor of Plastic Surgery, Vanderbilt University Medical Center
- Myranda Shirk, Data Science Institute, Vanderbilt University
Team Members:
- Catherine Dang, Undergraduate Student, Economics and Data Science
- Jaewoo Cho, Graduate Student, Data Science
- Hanlin Chen, Graduate Student, Data Science
- Lily Hofstetter, Undergraduate Student, Computer Science and Cognitive Science
- Jack Bijur, Undergraduate Student, Computer Science
This AI Deep Dive and the Cleft Palate Project exemplify our commitment to leveraging technology for social good, aiming to improve health outcomes and foster innovative research at Vanderbilt Data Science.
Are you interested in being a part of the research team? Contact us at datascience@vanderbilt.edu.