Gross motor ability predicts response to upper extremity rehabilitation in chronic stroke.

The majority of rehabilitation research focuses on the comparative effectiveness of different interventions in groups of patients, while much less is currently known regarding individual factors that predict response to rehabilitation. In a recent article, the authors presented a prognostic model to identify the sensorimotor characteristics predictive of the extent of motor recovery after Constraint-Induced Movement (CI) therapy amongst individuals with chronic mild-to-moderate motor deficit using the enhanced probabilistic neural network (EPNN). This follow-up paper examines which participant characteristics are robust predictors of rehabilitation response irrespective of the training modality. To accomplish this, EPNN was first applied to predict treatment response amongst individuals who received a virtual-reality gaming intervention (utilizing the same enrollment criteria as the prior study). The combinations of predictors that yield high predictive validity for both therapies, using their respective datasets, were then identified. High predictive classification accuracy was achieved for both the gaming (94.7%) and combined datasets (94.5%). Though CI therapy employed primarily fine-motor training tasks and the gaming intervention emphasized gross-motor practice, larger improvements in gross motor function were observed within both datasets. Poorer gross motor ability at pre-treatment predicted better rehabilitation response in both the gaming and combined datasets. The conclusion of this research is that for individuals with chronic mild-to-moderate upper extremity hemiparesis, residual deficits in gross motor function are highly responsive to motor restorative interventions, irrespective of the modality of training.

Computer-aided prediction of extent of motor recovery following constraint-induced movement therapy in chronic stroke.

Constraint-induced movement therapy (CI therapy) is a well-researched intervention for treatment of upper limb function. Overall, CI therapy yields clinically meaningful improvements in speed of task completion and greatly increases use of the more affected upper extremity for daily activities. However, individual improvements vary widely. It has been suggested that intrinsic feedback from somatosensation may influence motor recovery from CI therapy. To test this hypothesis, an enhanced probabilistic neural network (EPNN) prognostic computational model was developed to identify which baseline characteristics predict extent of motor recovery, as measured by the Wolf Motor Function Test (WMFT). Individual characteristics examined were: proprioceptive function via the brief kinesthesia test, tactile sensation via the Semmes-Weinstein touch monofilaments, motor performance captured via the 15 timed items of the Wolf Motor Function Test, stroke affected side. A highly accurate predictive classification was achieved (100% accuracy of EPNN based on available data), but facets of motor functioning alone were sufficient to predict outcome. Somatosensation, as quantified here, did not play a large role in determining the effectiveness of CI therapy.

Computer-Aided Diagnosis of Parkinson's Disease Using Enhanced Probabilistic Neural Network.

Early and accurate diagnosis of Parkinson's disease (PD) remains challenging. Neuropathological studies using brain bank specimens have estimated that a large percentages of clinical diagnoses of PD may be incorrect especially in the early stages. In this paper, a comprehensive computer model is presented for the diagnosis of PD based on motor, non-motor, and neuroimaging features using the recently-developed enhanced probabilistic neural network (EPNN). The model is tested for differentiating PD patients from those with scans without evidence of dopaminergic deficit (SWEDDs) using the Parkinson's Progression Markers Initiative (PPMI) database, an observational, multi-center study designed to identify PD biomarkers for diagnosis and disease progression. The results are compared to four other commonly-used machine learning algorithms: the probabilistic neural network (PNN), support vector machine (SVM), k-nearest neighbors (k-NN) algorithm, and classification tree (CT). The EPNN had the highest classification accuracy at 92.5% followed by the PNN (91.6%), k-NN (90.8%) and CT (90.2%). The EPNN exhibited an accuracy of 98.6% when classifying healthy control (HC) versus PD, higher than any previous studies.