Ambulatory Behavior Assessment Using Deep Learning.

Area of Science:

• Computer Vision
• Machine Learning
• Clinical Biomechanics

Background:

• Accurate quantification of patient ambulation is crucial for rehabilitation and clinical decision-making.
• Traditional methods for assessing ambulation can be subjective and labor-intensive.
• There is a need for objective, automated tools to monitor patient mobility in clinical settings.

Purpose of the Study:

• To develop and validate a deep neural network-based system for detecting people and assistive devices.
• To quantify diverse ambulatory activities and behaviors using computer vision and machine learning.
• To provide data for collaborative goal setting between clinicians and hospitalized patients.

Main Methods:

• A custom deep neural network object detection algorithm was implemented.
• The system detects individuals and assistive devices within clinical environments.
• Extracted features were used as input for machine learning models to quantify ambulation and its mode.

Main Results:

• The system successfully detected people and relevant assistive devices.
• Quantification of different ambulatory activities and related behaviors was achieved.
• The system accurately determined how a person ambulates and their mode of ambulation.

Conclusions:

• The developed system offers an objective method for assessing patient ambulation.
• This technology facilitates the creation, monitoring, and adjustment of ambulatory goals.
• It supports enhanced collaboration between clinicians and patients in managing mobility.