TUTORIALS

Tues June 18, 1030-1130am

 

Tutorial A

Title:
Leveraging Machine Learning Techniques to Unlock Gait Analysis in "TheWild"

Presenters:
James Carollo
Evelyn Cheng
Cameron Turner

Purpose:
This tutorial aims to equip attendees with the knowledge and practical skills needed to utilize open-source technology (OpenPose) for gait parameter quantification using video sources that are unconstrained and/or uncalibrated. This expands the measurement possibilities of routinely collected "observational video" and may facilitate lower-fidelity gait quantification outside the laboratory.

Target Audience:
Engineers, kinesiologists, researchers, and clinicians interested in expanding the reach of gait and motion analysis to the clinic, school, therapy gym, or playing field.

Summary:
OpenPose, an open-source machine learning/computer vision library, offers powerful capabilities for detecting and tracking human body keypoints, enabling researchers and practitioners to extract valuable gait metrics efficiently. Through live demonstration and in-depth discussion, participants will explore the intricacies of gait performance using OpenPose, including key concepts, data pre- and post-processing techniques, metric calculations, and interpretation of results. By the end of the session, attendees will have a comprehensive understanding of how they might leverage OpenPose effectively to enhance their gait analysis workflows.

Presentation Format:

  1. Introduction/Problem Statement (10 min)
    1. Observational video remains a fundamental tool in clinical movement analysis.
    2. Unconstrained and uncalibrated video can be collected anywhere.
    3. Benefits of calculating simple temporal/spatial parameters from video obtained in the field.
    4. Constraints imposed when relying on uncalibrated video sources.
    5. Goal: Demonstrate that open-source tools and machine learning techniques can be used to extract quantitative temporal/spatial data from uncalibrated video.

  2. Tutorial: Gait Analysis with OpenPose (35 min)
    1. Understanding OpenPose
      1. Overview of OpenPose technology
      2. Installation and setup
    2. Preprocessing data for gait analysis
      1. Data acquisition and preparation
      2. Quality control techniques
    3. Extracting coordinates with OpenPose
      1. Keypoint detection and tracking
      2. Parsing outputs
    4. Postprocessing model outputs
      1. Signal smoothing
      2. Calculating key gait parameters
      3. Comparison to ground truth
  3. Conclusions (5 min)
    1. Interpreting and Utilizing Gait Metrics
      1. Pitfalls
      2. Next steps
  4. Questions (10 min)

Learning Objectives:
By the end of this tutorial, participants will be able to:

  • Familiarize themselves with OpenPose technology and its capabilities for gait analysis.
  • Acquire skills in preprocessing gait data and ensuring data quality.
  • Apply OpenPose and postprocessing steps to extract meaningful gait metrics.
  • Knowledgeably discuss model interpretation and practical applications for research or clinical purposes.

 

Tutorial B

Title:
Innovations in Care: Integrating 3D Motion Analysis into a Two-Day Multidisciplinary Care Model for Pediatric Cerebral Palsy

Purpose:
To discuss the integration of quantitative 3D gait analysis into the development of the Cerebral Palsy (CP) Surgical Mobility Destination Center, part of the CP Program at Cincinnati Children's Hospital. This seminar will highlight the patient and family needs that led to the development of the program, as well as the technical challenges and resolutions that arose during program development.

Target Audience:
Engineers, kinesiologists, therapists, physicians, clinical support staff, and administrators interested in innovative models for multidisciplinary CP care.

Summary:
The mobility deficits associated with pediatric CP are generally managed across a spectrum of providers including physiatry, orthopaedic surgery, neurosurgery, and physical therapy. Quantitative 3D gait analysis can inform treatment options in all these areas, but availability of these specialists can preclude expeditious access to appropriate care. From our own experience, a child with CP seeking surgical tone or orthopedic care might require 6-9 months to complete a series of clinic visits. This poses even more of a challenge for patients who travel remotely.

In early 2020, providers in our institution's CP Program identified this prolonged and fragmented care model as a significant problem, delaying interventions and negatively impacting patient outcomes. They developed a program wherein a child with CP could consult with a multidisciplinary team, including a gait analysis in the Motion Analysis Lab. The program's overarching goal was to deliver a comprehensive assessment and recommendations, informed by clinical evaluation and quantitative gait analysis, within 48 hours or less. Over 90 patients have completed the program to date.

This seminar will present the challenges, creative solutions and important lessons that led to improved care delivery, patient family experience, and program growth.

Format:

Panel Presentation (40 minutes):

  1. Introduction (5 minutes)
  2. Specific Challenges (5 minutes)
    1. Limited understanding of multiple complex factors that contribute to impaired mobility in CP.
    2. Single specialist evaluation may only offer one treatment or surgical intervention familiar to that provider without considering other interventions or family goals/context.
    3. Navigating a large institution maze and ge_ng subspecialties to collaborate on a comprehensive plan has been challenging.
  3. C. Clinic Format and Specifics (10 minutes)
  4. Lessons and Adaptations (10 minutes)
    1. Stakeholder engagement
    2. Clinic triage algorithm
    3. Clinic format and shared mental model
    4. Post-operative therapy protocols to optimize outcomes
    5. Parental responses and support
  5. Outcomes (10 minutes)
    1. Clinic expansion
    2. Surgical conversion rates
    3. Patient family experience
    4. Mobility and quality of life outcomes

Sharing and Learning / Audience Feedback (20 minutes)

  1. Questions and comments from the group
  2. Ideas for leveraging successes across settings and providers
  3. Building collaboration

Learning Objectives:
At the conclusion of this presentation, participants will be able to:

  • Identify the limitng factors that can preclude expeditous patient care in a condition requiring multidisciplinary care.
  • Discuss the importance of patient intake and triaging processes in identifying patients appropriate for the program.
  • Discuss the particular technical challenges of integrating 3D Motion Analysis into this care model.