A Novel Framework for Quantifying Accuracy and Precision of Event Detection Algorithms in FES-Cycling

Area of Science:

• Rehabilitation Engineering
• Biomedical Signal Processing
• Neuroprosthetics

Background:

• Functional electrical stimulation (FES) aids movement in rehabilitation by electrically activating muscles.
• Current FES cycling relies on crank angle, which is sensitive to changes in seating position, causing desynchronized contractions.
• Adaptive algorithms offer potential for automatic FES pattern design by interpreting anatomical segments in real-time.

Purpose of the Study:

• To evaluate the accuracy and precision of three algorithms (Hilbert, BSgonio, GCI Observer) in detecting stimulation triggering events for FES cycling.
• To assess the adaptability of these algorithms for real-world applications under various conditions.
• To establish appropriate evaluation criteria for FES control algorithms.

Main Methods:

• Inertial data from passive cycling of six spinal cord injury (SCI) participants were analyzed.
• Three algorithms (Hilbert, BSgonio, GCI Observer) were tested against a linear phase reference baseline for event detection.
• Limits of Agreement (LoA) and Lin's Concordance Correlation Coefficient (CCC) were used to assess accuracy and precision over 780 events.

Main Results:

• The Hilbert and BSgonio algorithms met the validation criteria with LoA of +5.17/-6.34% and +2.25/-2.51%, respectively.
• The GCI Observer algorithm did not meet the criteria, showing LoA of +8.59/-27.89%.
• Normalizing detection delays to cycle duration proved effective for cadence-invariant evaluation.

Conclusions:

• Hilbert and BSgonio algorithms demonstrate sufficient accuracy and precision for adaptive FES cycling control.
• The GCI Observer requires further refinement for reliable event detection in FES applications.
• Standardized evaluation metrics, including normalized delays and CCC, are crucial for assessing FES control algorithm performance.