AI-Based Automatic System for Assessing Upper-Limb Spasticity of Patients With Stroke Through Voluntary Movement

Area of Science:

• Biomedical Engineering
• Rehabilitation Medicine
• Neurology

Background:

• Spasticity is a frequent complication following stroke, impacting voluntary movement.
• Limited research explores the connection between spasticity and volitional motor control.

Purpose of the Study:

• To introduce a novel automated system for quantifying upper-limb spasticity severity (SS) during voluntary movements.
• To assess spasticity in the elbow, wrist, thumb, and fingers.

Main Methods:

• A wearable system with 19 inertial measurement units and a pressure ball collected kinematic and force data.
• Participants performed four tasks: cone stacking (CS), fast flexion/extension (FFE), slow ball squeezing (SBS), and fast ball squeezing (FBS).
• Time/frequency domain features were extracted, selected using recursive feature elimination, and analyzed with five machine learning models.

Main Results:

• The cone stacking task showed high accuracy for elbow and finger spasticity assessment.
• The fast ball squeezing task was most effective for thumb and wrist spasticity.
• Fast ball squeezing emerged as the optimal task for assessing all four upper-limb joints.

Conclusions:

• The proposed automatic system accurately assesses upper-limb spasticity through voluntary movements.
• This represents a significant advancement in understanding the relationship between spasticity and volitional control.
• The system offers a promising tool for objective spasticity evaluation in stroke rehabilitation.