Accuracy of Step Count Estimations in Parkinson's Disease Can Be Predicted Using Ambulatory Monitoring.

Objectives: There are concerns regarding the accuracy of step count in Parkinson's disease (PD) when wearable sensors are used. In this study, it was predicted that providing the normal rhythmicity of walking was maintained, the autocorrelation function used to measure step count would provide relatively low errors in step count. Materials and Methods: A total of 21 normal walkers (10 without PD) and 27 abnormal walkers were videoed while wearing a sensor [Parkinson's KinetiGraph (PKG)]. Median step count error rates were observed to be <3% in normal walkers but ≥3% in abnormal walkers. The simultaneous accelerometry data and data from a 6-day PKG were examined and revealed that the 5th percentile of the spectral entropy distribution, among 10-s walking epochs (obtained separately), predicted whether subjects had low error rate on step count with reference to the manual step count from the video recording. Subjects with low error rates had lower Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS III) scores and UPDRS III Q10-14 scores than the high error rate counterparts who also had high freezing of gait scores (i.e., freezing of gait questionnaire). Results: Periods when walking occurred were identified in a 6-day PKG from 190 non-PD subjects aged over 60, and 155 people with PD were examined and the 5th percentile of the spectral entropy distribution, among 10-s walking epochs, was extracted. A total of 84% of controls and 72% of people with PD had low predicted error rates. People with PD with low bradykinesia scores (measured by the PKG) had step counts similar to controls, whereas those with high bradykinesia scores had step counts similar to those with high error rates. On subsequent PKGs, step counts increased when bradykinesia was reduced by treatment and decreased when bradykinesia increased. Among both control and people with PD, low error rates were associated with those who spent considerable time making walks of more than 1-min duration. Conclusion: Using a measure of the loss of rhythmicity in walking appears to be a useful method for detecting the likelihood of error in step count. Bradykinesia in subjects with low predicted error in their step count is related to overall step count but when the predicted error is high, the step count should be assessed with caution.

A method for measuring time spent in bradykinesia and dyskinesia in people with Parkinson's disease using an ambulatory monitor.

BACKGROUND: Fluctuations in motor function in Parkinson's Disease (PD) are frequent and cause significant disability. Frequently device assisted therapies are required to treat them. Currently, fluctuations are self-reported through diaries and history yet frequently people with PD do not accurately identify and report fluctuations. As the management of fluctuations and the outcomes of many clinical trials depend on accurately measuring fluctuations a means of objectively measuring time spent with bradykinesia or dyskinesia would be important. The aim of this study was to present a system that uses wearable sensors to measure the percentage of time that bradykinesia or dyskinesia scores are above a target as a means for assessing levels of treatment and fluctuations in PD. METHODS: Data in a database of 228 people with Parkinson's Disease and 157 control subjects, who had worn the Parkinson's Kinetigraph ((PKG, Global Kinetics Corporation™, Australia) and scores from the Unified Parkinson's Disease Rating Scale (UPDRS) and other clinic scales were used. The PKG's provided score for bradykinesia and dyskinesia every two minutes and these were compared to a previously established target range representing a UPDRS III score of 35. The proportion of these scores above target over the 6 days that the PKG was worn were used to derive the percent time in bradykinesia (PTB) and percent time in dyskinesia (PTD). As well, a previously describe algorithm for estimating the amplitude of the levodopa response was used to determine whether a subject was a fluctuator or non-fluctuator. RESULTS: Using this approach, a normal range of PTB and PTD based on Control subject was developed. The level of PTB and PTD experienced by people with PD was compared with their levels of fluctuation. There was a correlation (Pearson's ρ = 0.4) between UPDRS II scores and PTB: the correlation between Parkinson Disease Questionnaire scores and UPDRS Total scores and PTB and slightly lower. PTB and PTD fell in response to treatment for bradykinesia or dyskinesia (respectively) with greater sensitivity than clinical scales. CONCLUSIONS: This approach provides an objective assessment of the severity of fluctuations in Parkinson's Disease that could be used in in clinical trials and routine care.

Adaptive Learning for Robust Radial Basis Function Networks.

This article addresses the robust estimation of the output layer linear parameters in a radial basis function network (RBFN). A prominent method used to estimate the output layer parameters in an RBFN with the predetermined hidden layer parameters is the least-squares estimation, which is the maximum-likelihood (ML) solution in the specific case of the Gaussian noise. We highlight the connection between the ML estimation and minimizing the Kullback-Leibler (KL) divergence between the actual noise distribution and the assumed Gaussian noise. Based on this connection, a method is proposed using a variant of a generalized KL divergence, which is known to be more robust to outliers in the pattern recognition and machine-learning problems. The proposed approach produces a surrogate-likelihood function, which is robust in the sense that it is adaptive to a broader class of noise distributions. Several signal processing experiments are conducted using artificially generated and real-world data. It is shown that in all cases, the proposed adaptive learning algorithm outperforms the standard approaches in terms of mean-squared error (MSE). Using the relative increase in the MSE for different noise conditions, we compare the robustness of our proposed algorithm with the existing methods for robust RBFN training and show that our method results in overall improvement in terms of absolute MSE values and consistency.

Sparse Principal Component Analysis With Preserved Sparsity Pattern.

Principal component analysis (PCA) is widely used for feature extraction and dimension reduction in pattern recognition and data analysis. Despite its popularity, the reduced dimension obtained from the PCA is difficult to interpret due to the dense structure of principal loading vectors. To address this issue, several methods have been proposed for sparse PCA, all of which estimate loading vectors with few non-zero elements. However, when more than one principal component is estimated, the associated loading vectors do not possess the same sparsity pattern. Therefore, it becomes difficult to determine a small subset of variables from the original feature space that have the highest contribution in the principal components. To address this issue, an adaptive block sparse PCA method is proposed. The proposed method is guaranteed to obtain the same sparsity pattern across all principal components. Experiments show that applying the proposed sparse PCA method can help improve the performance of feature selection for image processing applications. We further demonstrate that our proposed sparse PCA method can be used to improve the performance of blind source separation for functional magnetic resonance imaging data.

Consistent Estimation of Dimensionality for Data-Driven Methods in fMRI Analysis.

Data-driven methods, such as principal component analysis and independentcomponent analysis, have been successfully applied to functionalmagnetic resonance imaging (fMRI) data in particular and neuro-imaging data in general. A central issue of thesemethods is the importance of correctly selecting the number of components to be used in the factor model. This issue is often addressed using a model selection criterion, where the goodness-of-fit term is obtained from the log-likelihood function. In this paper, an alternative criterion is proposed for selecting the number of components. Unlike existingmodel selection criteria that use the log-likelihood function, the proposed goodness-of-fit termuses the sum of squares of the smallest eigenvalues of the sample covariance matrix. The proposed criterion is obtained from the asymptotic distribution of the goodness-of-fit term, for which consistency is established. This criterion has a straight-forward implementation and is shown to outperform conventional model selection criteria used in fMRI data analysis. Experiments are conducted using simulated and real fMRI data, in which improved performance is obtained by the proposed criterion, both in terms of accuracy and consistency under data variabilities.

Analysis of human scream and its impact on text-independent speaker verification.

Scream is defined as sustained, high-energy vocalizations that lack phonological structure. Lack of phonological structure is how scream is identified from other forms of loud vocalization, such as "yell." This study investigates the acoustic aspects of screams and addresses those that are known to prevent standard speaker identification systems from recognizing the identity of screaming speakers. It is well established that speaker variability due to changes in vocal effort and Lombard effect contribute to degraded performance in automatic speech systems (i.e., speech recognition, speaker identification, diarization, etc.). However, previous research in the general area of speaker variability has concentrated on human speech production, whereas less is known about non-speech vocalizations. The UT-NonSpeech corpus is developed here to investigate speaker verification from scream samples. This study considers a detailed analysis in terms of fundamental frequency, spectral peak shift, frame energy distribution, and spectral tilt. It is shown that traditional speaker recognition based on the Gaussian mixture models-universal background model framework is unreliable when evaluated with screams.