Adaptive Recalibration of Prosthetic Leg Neural Control System
Overview
- Phase
- Not Applicable
- Intervention
- Not specified
- Conditions
- Amputation
- Sponsor
- Shirley Ryan AbilityLab
- Enrollment
- 22
- Locations
- 1
- Primary Endpoint
- Decreased error rates for pattern recognition system used to predict ambulation modes
- Status
- Completed
- Last Updated
- 5 years ago
Overview
Brief Summary
The purpose of this study is for transfemoral amputees to walk with an experimental robotic prosthesis. Electric signals will be measured from their muscles and used to help control an artificial leg. The investigators will record from sensors placed on a prosthesis and electric signals measured from muscles in the participants leg to see if the investigators can develop better computer programs to help predict subject actions and prostheses function.
Detailed Description
The investigators propose to use a powered knee-ankle prosthesis that is not yet commercially available. The hierarchical control framework the investigators are developing will be equally applicable to any prosthetic leg that needs to be transitioned between ambulation modes, including microprocessor-controlled passive devices. The overall objective is to develop and evaluate an adaptive framework for controlling lower limb prostheses that compensates for changes in EMG signals. When a participant walks on a lower limb prosthesis, the output of the high-level controller (or ambulation mode predictor) directly influences patterns generated by the participant. After the participant has completed the subsequent stride, a gait pattern estimator (GPE), will provide a label of what the participant actually did. This may differ from the ambulation mode predictor output if there was a misclassification. The label will then be used to update the ambulation mode predictor algorithm such that future steps are predicted with higher accuracies. Finally, the resulting system will be transferred to an embedded system and tested in real-time with transfemoral amputees and compared to a non-adaptive system.
Investigators
Levi Hargrove
Director, Neural Engineering for Prosthetics and Orthotics Laboratory
Shirley Ryan AbilityLab
Eligibility Criteria
Inclusion Criteria
- •Lower Limb Amputees
- •K2/K3/K4 ambulators
Exclusion Criteria
- •Over 250lbs body weight
- •Inactive, physically unfit
- •cognitive deficits or visual impairment that would impair their ability to give informed consent or to follow simple instructions during the experiments
- •Pregnant women
- •co-morbidity that interferes with the study (e.g. stroke, pace maker placement, severe ischemia cardiac disease, etc.)
- •Able-bodied Subjects:
- •Inclusion Criteria:
- •no injury on either lower extremity
- •Exclusion Criteria:
- •inactive, physically unfit
Outcomes
Primary Outcomes
Decreased error rates for pattern recognition system used to predict ambulation modes
Time Frame: Assessed at approximately 2 months and 6 months after enrollment
Pattern recognition algorithms have been used to allow seamless and automatic transitioning between ambulation modes. Classification errors result in the prosthesis predicting the wrong ambulation mode. A decrease in errors results in improved mode prediction by the prosthesis. EMG from the participant and mechanical sensor data from the prosthesis are processed with the use of a phase-based-dependent pattern recognition classification method. The data collection will yield three groups of 10 real-time trials. The investigators primary analysis will be a repeated measures ANOVA with a planned contrast between the groups. The investigators will also complete a secondary analysis using the data collected while the participants ambulated outside of the laboratory. The total number of misclassifications will be computed. This will allow the investigator to evaluate the rate at which the overall classification system adapted.