Intelligent data analysis of instrumented gait data in stroke patients-a systematic review

Instrumented gait analysis (GA) may be used to analyze the causes of gait deviation in stroke patients but generates a large amount of complex data. The task of transforming this data into a comprehensible report is cumbersome. Intelligent data analysis (IDA) refers to the use of computational methods in order to analyze quantitative data more effectively. The purpose of this review was to identify and appraise the available IDA methods for handling GA data collected from patients with stroke using the standard equipment of a gait lab (3D/2D motion capture, force plates, EMG). Eleven databases were systematically searched and fifteen studies that employed some type of IDA method for the analysis of kinematic and/or kinetic and/or EMG data in populations involving stroke patients were identified. Four categories of IDA methods were employed for the analysis of sensor-acquired data in these fifteen studies: classification methods, dimensionality reduction methods, clustering methods and expert systems. The methodological quality of these studies was critically appraised by examining sample characteristics, measurements and IDA properties. Three overall methodological shortcomings were identified: (1) small sample sizes and underreported patient characteristics, (2) testing of which method is best suited to the analysis was neglected and (3) lack of stringent validation procedures. No IDA method for GA data from stroke patients was identified that can be directly applied to clinical practice. Our findings suggest that the potential provided by IDA methods is not being fully exploited.