Ankle Exoskeleton for Stroke Gait Enhancement

Temporal Convolutional Network (TCN) model performance (Joint moment estimation accuracy)

This outcome represents the error with which the deep learning model embedded into our ankle exoskeleton's microprocessor predicts ankle joint moments in stroke patients. Specifically, the coefficient of determination (R²) is computed between the predicted ankle joint moments and the ground truth measurements. Ground truth measurements are obtained from a laboratory-grade force plate system and inverse dynamics calculations. Ankle joint moment predictions are made at a frequency of 200 Hz and compared to the laboratory-measured values. For these measures, higher R² values (closer to 1.0) indicate better correlation between predicted and actual ankle joint moments. This metric provides a comprehensive assessment of the exoskeleton's ability to accurately estimate ankle joint moments in stroke patients during tasks, with improved outcomes representing better assistive capabilities for the user.

Metabolic cost for level ground walking

Metabolic energy expenditure will be quantified using an indirect calorimetry system (Parvo Medics, UT) that measures oxygen consumption (VO₂) and carbon dioxide production (VCO₂) during experimental tasks. Measurements will be collected from each participant during a 5-minute baseline standing period followed by level ground walking trials under two conditions: without the exoskeleton, with the exoskeleton in a powered state. Metabolic cost will be calculated from respiratory gas exchange data (VO₂ and VCO₂) using Brockway equations \[1\] for energy expenditure. Comparisons between the two conditions will be conducted to assess the effectiveness of the exoskeleton with respect to metabolic cost. Energy expenditure (kilojoule/minute) = 16.58 VO₂ (Liters/minute) +4.51VCO₂ (Liters/minute) \[1\] Brockway, J. M. "Derivation of formulae used to calculate energy expenditure in man." Human nutrition. Clinical nutrition 41.6 (1987): 463-471.

Biological Joint Work

Mechanical work performed by the lower limb joints will be quantified through biomechanical analysis of motion capture data. Joint moments and angular velocities will be derived through inverse dynamics and kinematics, respectively. Joint power, calculated as the product of joint moment and angular velocity, will be integrated with respect to time using trapezoidal integration to determine mechanical work. Positive and negative work will be calculated by separately integrating positive and negative joint powers, providing comprehensive quantification of joint energy generation and absorption at each joint during the movement tasks.

Single limb stance time asymmetry (temporal)

This will be measured as the participant walks across a gait mat and/or via motion capture as the time spent on the right and left leg is calculated. The index will be calculated as the difference between the time spent in single-limb support for the right and left legs during walking and expressed as a percentage with a value of 0 indicating perfect symmetry and greater values indicating larger asymmetry.

Step length asymmetry (spatial)

This will be measured as the participant walks across a gait mat and/or via motion capture as the distance traversed by the right and left leg for each step. The index will be calculated as the difference between the step lengths of the right and left legs during walking and expressed as a percentage with a value of 0 indicating perfect symmetry and greater values indicating larger asymmetry.

Interlimb propulsion asymmetry (kinetic)

This will be measured as the participant walks on an instrumented treadmill with force plates, by calculating the peak anterior-posterior propulsive impulse generated by the paretic and non-paretic limbs during late stance. A propulsion symmetry index will then be calculated using the difference between the propulsive impulses of the two limbs, expressed as a percentage. A value of 0 indicates perfect symmetry, while greater values reflect larger propulsion asymmetry between limbs, with lower paretic propulsion indicating impaired forward progression capacity.

Trailing limb angle (kinematic)

This will be measured as the participant walks on an instrumented treadmill via motion capture, by calculating the sagittal-plane angle between the vertical axis and the line connecting the hip joint center to the ankle joint center at the moment of contralateral initial contact. Values will be reported in degrees, with larger angles indicating greater trailing-limb extension and push-off capacity during gait.