Computer-aided identification of stroke-associated motor impairments using a virtual reality augmented robotic system

Akbarifar, Faranak; Dukelow, Sean P.; Mousavi, Parvin; Scott, Stephen H.

doi:10.1080/21681163.2021.2009377

Cited by 2 publications

(5 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the TIA dataset, the notable contribution of speed-related features to classification as "impaired" suggests that participants in this group perform their tasks at a significantly slower rate (Figure 4). Features like "no movement end" and "end target not reached" in both test sets aligns with our previous study, which identified such difficulties as a reason for subclustering within various CMSA levels of patients [19]. Furthermore, it is worth noting that features from both arms of VGR are instrumental, justifying the inclusion of features from both arms.…”

Section: Discussionsupporting

confidence: 87%

“…Middle rows of Table 2 exclusively present the performance of the top-performing models from the previous table, focusing on the least impaired stroke participants (CMSA Scores = 7 for both arms) and the control group. Notably, MLP-evidential method outperforms the others in this specific context, achieving an increase in sensitivity to 0.55 compared to our previous study [19].…”

Section: Binary Classification Results On Stroke/control Test Setcontrasting

confidence: 43%

“…According to what was observed in our previous research [19], behavioral assessment using robot-based measurement technologies provide a wealth of objective information. This information proves highly valuable, as it can be effectively harnessed by both shallow and deep machine learning methods to effectively differentiate between the majority of stroke participants, including those with minimal impairment, and control participants (refer to Table 2).…”

Section: Discussionmentioning

confidence: 78%

“…Their reported 94% accuracy using a simple artificial neural network exceeds our average but lacks sensitivity and specificity details, nor does it cover detecting impairments in minimally impaired stroke participants. Contrarily, [19] employed deep learning, providing insights into impairments, yet our study, leveraging deep neural networks with evidential losses, excels in detecting subtle impairments and estimating uncertainty. Our approach aids clinicians by identifying samples for further examination, enhancing stroke detection sensitivity.…”

Section: Discussionmentioning

confidence: 99%

“…Notably in [18], a number of classical machine learning techniques, as well as an artificial neural network, have shown promise in differentiating stroke participants from control groups based on the kinematic features acquired from a robotic reaching task. A more recent study [19] utilized an autoencoder-based classifier to distinguish least impaired (CMSA = 7, scored as normal) stroke participants from healthy controls, with an AUC equal to 0.84, yet room for improvement in classification sensitivity remained.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Optimizing Stroke Detection Using Evidential Networks and Uncertainty-Based Refinement

Akbarifar,

Dukelow,

Jin

et al. 2024

Preprint

Self Cite

View full text Add to dashboard Cite

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.

show abstract

Section: Discussionsupporting

confidence: 87%

Section: Binary Classification Results On Stroke/control Test Setcontrasting

confidence: 43%

Section: Discussionmentioning

confidence: 78%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations