VST: Virtual Speech Therapist

Virtual Speech Therapist for people with Parkinson disease
The advent of computer-based learning systems that combine speech processing, computer vision and lifelike animated computer characters provides a new opportunity to adapt clinician-directed treatment to patient home self-training, while preserving feedback and clinician monitoring (Buyukkokten et al., 2000). These technologies promise increased patient access to treatment to sustain treatment effects and to enhance treatment outcomes, and offer patients the opportunity to receive treatment early in the disease process, which may have preventative value. In addition, these animated learning systems have the capability to acquire data in “real world” clinical and home settings to generate large patient sample sizes for effectiveness studies to support reimbursement as well as to monitor quality of treatment worldwide.

The proposed work will address two specific aims:
1. Develop a computer-based training program using the Lee Silverman Voice Treatment (LSVT) in which a virtual therapist monitors and evaluates the verbal and visual behaviors of patients during LSVT sessions, and responds appropriately to patient behaviors by modeling the auditory and visual behaviors of expert LSVT clinicians in an engaging and effective manner; and
2. Explore the feasibility of using the virtual therapist as either a substitute or an effective adjunct to LSVT training in clinical settings.

Aim 1 will generate a complete working system that will interact with patients, will log, analyze and display all patient and system behaviors, and will playback sessions for review and additional analyses. The work will also produce an API (Application Programmers Interface, or authoring tools) for developing systems in new languages. The system will be developed using participatory design methodology, which involves close collaboration between the PI, co-PI, consultants, software developers, end users (LSVT patients who have completed LSVT training: LSVT veterans), and other stakeholders at all stages of the design and test process.

Aim 2 will generate data to explore the “real world” effectiveness of voice treatment for individuals with IPD using a multilingual virtual therapist. Findings from this work have the potential to create a systematic, worldwide network of voice treatment sites and facilitate acquisition of multi-site voice treatment effectiveness data.

These studies provide an excellent opportunity to break new ground in our goal to improve communication and the quality of life of individuals with IPD. Demonstrating feasibility in this domain is expected to lead to more ambitious and comprehensive efforts to evaluate efficacy and could lead to future development of accessible, inexpensive and effective treatment opportunities for individuals across a wide range of communication disorders.