A pilot on the use of stride cadence for the characterization of walking ability in lower limb amputees

Abstract

Amputations are a prominent affliction that occur worldwide, with causes ranging from congenital, disease-based, or external reasons such as trauma. Prosthesis provides the closest alternative functional replacement to the loss of a limb. Before any form of rehabilitation support can be offered to amputee patients, an assessment of their degree and level of mobility first needs to be evaluated using the K-level grading system. The typical means towards the assigning of a K-level grading is through qualitative methods, which have been criticized for being subjective and, at times, imprecise. As a means towards remedying this shortcoming, we investigated the prospect of utilizing data from wearable sensors for analyzing the stride pattern and cadence of various subjects towards the quantitative inference of a K-level. This was accomplished using data from accelerometers, alongside advanced signal processing and machine learning models, towards the quantitative identification and differentiation of the various K-levels of amputees of varied levels of mobility. The experimental results showed that this aim could be accomplished under the circumstance investigated and the models applied as part of this research. Additional analysis was also done on the use of data from accelerometers towards the differentiation between amputated and non-amputated subjects, which showed that the cohorts could be classified and differentiated using purely accelerometer data and the accompanying postprocessing methods.

Keywords

• Lower limb
• Signal processing
• Machine learning
• Artificial intelligence
• LSDL
• Prosthesis