Optimizing Stroke Detection Using Evidential Networks and Uncertainty-Based Refinement

Abstract Background:Technologies such as interactive robotics and motion capture systems permit the development of kinematic-based approaches to assess motor impairments in stroke survivors. Here we utilise the Kinarm Exoskeleton robotic system and deep learning techniques to explore differences in motor performance between healthy controls, individuals with stroke and transient ischemic attacks (TIA).Methods:Building upon previous research that employed deep learning methods to distinguish between minimally impaired stroke patients and healthy controls using Kinarm data, this study introduces a novel dimension by estimating the confidence or uncertainty of the model's predictions. An evidential network is employed to measure this confidence, which subsequently aids in the refinement of training and testing datasets.Results:The application of deep learning techniques in this context proves to be promising. By utilizing uncertainty measures to systematically enhance datasets, the sensitivity and specificity of detecting stroke-related impairments are improved. Furthermore, this model is extended to address the detection of potential impairments in individuals following TIA, where traditional methods often fall short. The hypothesis that the deep learning model has the capacity to detect impairment is tested, with initial results indicating its potential in identifying impairments in individuals with TIA based on subtle but measurable motor deficits.Conclusions:This comprehensive investigation highlights the value of deep learning in the assessment of neurological conditions using Kinarm. The introduced uncertainty estimation offers a nuanced approach to data refinement, enhancing the clinical utility of stroke detection and expanding to identification of potential impairments following TIA.