Study rationale

Dystonia, a neurological condition causing involuntary muscle contractions, significantly impairs the quality of life through pain and disability. Botulinum toxin (BoNT) injections are a first-line treatment for this condition, but their effectiveness heavily depends on the expertise of specialized neurologists. Many general practitioners lack the detailed knowledge required to optimize muscle selection and dosing, leading to treatment gaps and healthcare access issues. This study addresses that gap by leveraging AI to democratize BoNT administration for cervical dystonia treatment.

hypothesis

Computer vision AI can enable general practitioners to effectively administer BoNT injections, a skill currently limited to specialists. Specifically, the AI model will analyze muscle selection and dosing patterns to predict the outcomes of BoNT injections for patients with cervical dystonia.

study design

The study will involve 50 patients with cervical dystonia, capturing standardized photos of their posture before and after BoNT injections using the in-house VisionMD software. The study has two primary aims:

  • Aim 1: Develop an AI model to predict the outcome of muscle injection patterns in treating cervical dystonia. Using photos and movement disorder specialists’ injection data, the model will assess improvements in patients’ posture based on angles between anatomical landmarks.
  • Aim 2: Characterize the relationship between toxin dosage and changes in dystonia posture using AI. The model will analyze the correlation between dosage levels and muscle response to improve dystonic posture angles.

impact on dystonia treatment

This project has the potential to revolutionize dystonia treatment by making specialized knowledge accessible to general practitioners through AI. The AI platform could reduce the current reliance on highly specialized movement disorder neurologists, expanding access to effective BoNT therapy and improving patient outcomes worldwide.

next steps for development

The data generated from this pilot project will form the basis for a larger NIH R01 proposal, aiming to refine the AI model and expand its application to other types of commercially available BoNT treatments. Future developments could further enhance the AI’s accuracy in muscle selection and dosage personalization, optimizing dystonia treatment protocols.

other relevant information

  • Innovative Approach: By using AI and computer vision, this project stands out as a cutting-edge solution for improving the precision of BoNT injections, enhancing both patient care and healthcare accessibility.
  • Pilot Funding: This one-year pilot study is supported by a grant of $39,460, which will cover patient recruitment, data collection, AI model development, and software enhancements.

study team

  • Joshua Wong, MD –Dr. Wong is an assistant professor in the Department of Neurology. He has a clinical and research focus in deep brain stimulation (DBS) for movement disorders and other neurologic diseases. The Wong Lab conducts advanced neuroimaging analyses through connectomics and computational modeling along with electrophysiologic analyses of neuronal local field potentials to study structural and functional correlates of neurodegenerative diseases. Dr. Wong engages in multi-disciplinary collaborations to apply machine learning techniques to multi-modal datasets in order to optimize DBS therapy.  
  • Diego Guarin, PhD – Dr. Guarin is an assistant professor in the Department of Applied Physiology and Kinesiology with expertise in computer vision AI, particularly video-based markerless kinematics. He will provide insight into model development and analysis of photographic data, and support for the in-house automated photo capture software, VisionMD. Developed by the Guarin Lab, VisionMD captures kinematic information from patients with movement disorders.
  • Adolfo Ramirez-Zamora, MD – Dr. Ramirez is a professor in the Department of Neurology with expertise in neuromodulation and BoNT injections for movement disorders, particularly cervical dystonia. He will provide guidance on the clinical aspects of AI model development and offer insights into the clinical interpretation of AI outputs. Additionally, he will facilitate patient recruitment for this study.