Characteristics of muscle activation patterns at the ankle in stroke patients during walking.

Abstract

Stroke causes impairment of the sensory and motor systems; this can lead to difficulties in walking and participation in society. For effective rehabilitation it is important to measure the essential characteristics of impairment and associate these with the nature of disability. Efficient gait requires a complex interplay of muscles. Surface electromyography(sEMG) can be used to measure muscle activity and to observe disruption to this interplay after stroke. Yet, classification of this disruption in stroke patients has not been achieved. It is hypothesised that features identified from the sEMG signal can be used to classify underlying impairments. A clinically viable gait analysis system has been developed, integrating an in-house wireless sEMG system synchronised with bilateral video and inertial orientation sensors. Signal processing techniques have been extended and implemented, appropriate for use with sEMG. These techniques have focussed on frequency domain features using wavelet analysis and muscle activation patterns using principal component analysis. The system has been used to measure gait from stroke patients and un-impaired subjects. Characteristic patterns of activity from the ankle musculature were defined using principal component analysis of the linear envelope. Patients with common patterns of tibialis anterior activity did not necessarily share common patterns of gastrocnemius or soleus activity. Patients with similar linear envelope patterns did not always present with the same kinematic profiles. The relationship between observable impairments, kinematics and sEMG is seen to be complex and there is therefore a need for a multidimensional view of gait data in relation to stroke impairment. The analysis of instantaneous mean frequency and time-frequency has revealed additional periods of activity not obvious in the linear or raw signal representation. Furthermore, characteristic calf activity was identified that may relate to abnormal reflex activity. This has provided additional information with which to group characteristic muscle activity. An evaluation of the co-activation of gastrocnemius and tibialis anterior muscles using a sub-band filtering technique revealed three groups; those with distinct co-activation, those with little co-activation and those with continuous activity in the antagonistic pair across the stride. Signal features have been identified in sEMG recordings from stroke patients whilst walking extending current signal processing techniques. Common features of the sEMG and movement have been grouped creating a decision matrix. These results have contributed to the field of clinical measurement and diagnosis because interpretation of this decision matrix is related to underlying impairment. This has provided a framework from which subsequent studies can classify characteristic patterns of impairment within the stroke population; and thus assist in the provision of rehabilitative interventions.