Ensembling crowdsourced seizure prediction algorithms using long-term human intracranial EEG.

Seizure prediction is feasible, but greater accuracy is needed to make seizure prediction clinically viable across a large group of patients. Recent work crowdsourced state-of-the-art prediction algorithms in a worldwide competition, yielding improvements in seizure prediction performance for patients whose seizures were previously found hard to anticipate. The aim of the current analysis was to explore potential performance improvements using an ensemble of the top competition algorithms. The results suggest that minor increments in performance may be possible; however, the outcomes of statistical testing limit the confidence in these increments. Our results suggest that for the specific algorithms, evaluation framework, and data considered here, incremental improvements are achievable but there may be upper bounds on machine learning-based seizure prediction performance for some patients whose seizures are challenging to predict. Other more tailored approaches that, for example, take into account a deeper understanding of preictal mechanisms, patient-specific sleep-wake rhythms, or novel measurement approaches, may still offer further gains for these types of patients.

Epilepsyecosystem.org: crowd-sourcing reproducible seizure prediction with long-term human intracranial EEG.

Accurate seizure prediction will transform epilepsy management by offering warnings to patients or triggering interventions. However, state-of-the-art algorithm design relies on accessing adequate long-term data. Crowd-sourcing ecosystems leverage quality data to enable cost-effective, rapid development of predictive algorithms. A crowd-sourcing ecosystem for seizure prediction is presented involving an international competition, a follow-up held-out data evaluation, and an online platform, Epilepsyecosystem.org, for yielding further improvements in prediction performance. Crowd-sourced algorithms were obtained via the 'Melbourne-University AES-MathWorks-NIH Seizure Prediction Challenge' conducted at kaggle.com. Long-term continuous intracranial electroencephalography (iEEG) data (442 days of recordings and 211 lead seizures per patient) from prediction-resistant patients who had the lowest seizure prediction performances from the NeuroVista Seizure Advisory System clinical trial were analysed. Contestants (646 individuals in 478 teams) from around the world developed algorithms to distinguish between 10-min inter-seizure versus pre-seizure data clips. Over 10 000 algorithms were submitted. The top algorithms as determined by using the contest data were evaluated on a much larger held-out dataset. The data and top algorithms are available online for further investigation and development. The top performing contest entry scored 0.81 area under the classification curve. The performance reduced by only 6.7% on held-out data. Many other teams also showed high prediction reproducibility. Pseudo-prospective evaluation demonstrated that many algorithms, when used alone or weighted by circadian information, performed better than the benchmarks, including an average increase in sensitivity of 1.9 times the original clinical trial sensitivity for matched time in warning. These results indicate that clinically-relevant seizure prediction is possible in a wider range of patients than previously thought possible. Moreover, different algorithms performed best for different patients, supporting the use of patient-specific algorithms and long-term monitoring. The crowd-sourcing ecosystem for seizure prediction will enable further worldwide community study of the data to yield greater improvements in prediction performance by way of competition, collaboration and synergism.10.1093/brain/awy210_video1awy210media15817489051001.

Effect of sensory and motor connectivity on hand function in pediatric hemiplegia.

OBJECTIVE: We tested the hypothesis that somatosensory system injury would more strongly affect movement than motor system injury in children with unilateral cerebral palsy (USCP). This hypothesis was based on how somatosensory and corticospinal circuits adapt to injury during development; whereas the motor system can maintain connections to the impaired hand from the uninjured hemisphere, this does not occur in the somatosensory system. As a corollary, cortical injury strongly impairs sensory function, so we hypothesized that cortical lesions would impair hand function more than subcortical lesions. METHODS: Twenty-four children with unilateral cerebral palsy had physiological and anatomical measures of the motor and somatosensory systems and lesion classification. Motor physiology was performed with transcranial magnetic stimulation and somatosensory physiology with vibration-evoked electroencephalographic potentials. Tractography of the corticospinal tract and the medial lemniscus was performed with diffusion tensor imaging, and lesions were classified by magnetic resonance imaging. Anatomical and physiological results were correlated with measures of hand function using 2 independent statistical methods. RESULTS: Children with disruptions in the somatosensory connectivity and cortical lesions had the most severe upper extremity impairments, particularly somatosensory function. Motor system connectivity was significantly correlated with bimanual function, but not unimanual function or somatosensory function. INTERPRETATION: Both sensory and motor connectivity impact hand function in children with USCP. Somatosensory connectivity could be an important target for recovery of hand function in children with USCP. Ann Neurol 2017;82:766-780.

An automated, electronic assessment tool can accurately classify older adult postural stability.

Current methods of balance assessment in the clinical environment are often subjective, time-consuming and lack clinical relevance for non-ambulatory older adults. The objective of this study was to develop a novel method of balance assessment that utilizes data collected using the Microsoft Kinect 2 to create a Berg Balance Scale score, which is completely determined by statistical methods rather than by human evaluators. 74 older adults, both healthy and balance impaired, were recruited for this trial. All participants completed the Berg Balance Scale (BBS) which was scored independently by trained physical therapists. Participants then completed the items of the "Modified Berg Balance Scale" in front of the Microsoft Kinect camera. Kinematic data collected during this measurement was used to train a feed-forward neural network that was used to assign a Berg Balance Scale score. The neural network model estimated the clinician-assigned BBS score to within a median of 0.93 points for the participants in our sample population (range: 0.02-5.69). Using low-cost depth sensing camera technology and a clinical protocol that takes less than 5 min to complete in both ambulatory and non-ambulatory older adults, the method outlined in this manuscript can accurately predict a participant's BBS score and thereby identify whether they are deemed a high fall risk or not. If implemented correctly, this could enable fall prevention services to be deployed in a timely fashion using low-cost, accessible technology, resulting in improved safety of older adults.

MOABB: trustworthy algorithm benchmarking for BCIs.

OBJECTIVE: Brain-computer interface (BCI) algorithm development has long been hampered by two major issues: small sample sets and a lack of reproducibility. We offer a solution to both of these problems via a software suite that streamlines both the issues of finding and preprocessing data in a reliable manner, as well as that of using a consistent interface for machine learning methods. APPROACH: By building on recent advances in software for signal analysis implemented in the MNE toolkit, and the unified framework for machine learning offered by the scikit-learn project, we offer a system that can improve BCI algorithm development. This system is fully open-source under the BSD licence and available at https://github.com/NeuroTechX/moabb. MAIN RESULTS: We analyze a set of state-of-the-art decoding algorithms across 12 open access datasets, including over 250 subjects. Our results show that even for the best methods, there are datasets which do not show significant improvements, and further that many previously validated methods do not generalize well outside the datasets they were tested on. SIGNIFICANCE: Our analysis confirms that BCI algorithms validated on single datasets are not representative, highlighting the need for more robust validation in the machine learning for BCIs community.

Using Data From the Microsoft Kinect 2 to Quantify Upper Limb Behavior: A Feasibility Study.

The objective of this study was to assess whether the novel application of a machine learning approach to data collected from the Microsoft Kinect 2 (MK2) could be used to classify differing levels of upper limb impairment. Twenty-four healthy subjects completed items of the Wolf Motor Function Test (WMFT), which is a clinically validated metric of upper limb function for stroke survivors. Subjects completed the WMFT three times: 1) as a healthy individual; 2) emulating mild impairment; and 3) emulating moderate impairment. A MK2 was positioned in front of participants, and collected kinematic data as they completed the WMFT. A classification framework, based on Riemannian geometry and the use of covariance matrices as feature representation of the MK2 data, was developed for these data, and its ability to successfully classify subjects as either "healthy," "mildly impaired," or "moderately impaired" was assessed. Mean accuracy for our classifier was 91.7%, with a specific accuracy breakdown of 100%, 83.3%, and 91.7% for the "healthy," "mildly impaired," and "moderately impaired" conditions, respectively. We conclude that data from the MK2 is of sufficient quality to perform objective motor behavior classification in individuals with upper limb impairment. The data collection and analysis framework that we have developed has the potential to disrupt the field of clinical assessment. Future studies will focus on validating this protocol on large populations of individuals with actual upper limb impairments in order to create a toolkit that is clinically validated and available to the clinical community.

Using data from the Microsoft Kinect 2 to determine postural stability in healthy subjects: A feasibility trial.

The objective of this study was to determine whether kinematic data collected by the Microsoft Kinect 2 (MK2) could be used to quantify postural stability in healthy subjects. Twelve subjects were recruited for the project, and were instructed to perform a sequence of simple postural stability tasks. The movement sequence was performed as subjects were seated on top of a force platform, and the MK2 was positioned in front of them. This sequence of tasks was performed by each subject under three different postural conditions: "both feet on the ground" (1), "One foot off the ground" (2), and "both feet off the ground" (3). We compared force platform and MK2 data to quantify the degree to which the MK2 was returning reliable data across subjects. We then applied a novel machine-learning paradigm to the MK2 data in order to determine the extent to which data from the MK2 could be used to reliably classify different postural conditions. Our initial comparison of force plate and MK2 data showed a strong agreement between the two devices, with strong Pearson correlations between the trunk centroids "Spine_Mid" (0.85 ± 0.06), "Neck" (0.86 ± 0.07) and "Head" (0.87 ± 0.07), and the center of pressure centroid inferred by the force platform. Mean accuracy for the machine learning classifier from MK2 was 97.0%, with a specific classification accuracy breakdown of 90.9%, 100%, and 100% for conditions 1 through 3, respectively. Mean accuracy for the machine learning classifier derived from the force platform data was lower at 84.4%. We conclude that data from the MK2 has sufficient information content to allow us to classify sequences of tasks being performed under different levels of postural stability. Future studies will focus on validating this protocol on large populations of individuals with actual balance impairments in order to create a toolkit that is clinically validated and available to the medical community.

Approximate joint diagonalization and geometric mean of symmetric positive definite matrices.

We explore the connection between two problems that have arisen independently in the signal processing and related fields: the estimation of the geometric mean of a set of symmetric positive definite (SPD) matrices and their approximate joint diagonalization (AJD). Today there is a considerable interest in estimating the geometric mean of a SPD matrix set in the manifold of SPD matrices endowed with the Fisher information metric. The resulting mean has several important invariance properties and has proven very useful in diverse engineering applications such as biomedical and image data processing. While for two SPD matrices the mean has an algebraic closed form solution, for a set of more than two SPD matrices it can only be estimated by iterative algorithms. However, none of the existing iterative algorithms feature at the same time fast convergence, low computational complexity per iteration and guarantee of convergence. For this reason, recently other definitions of geometric mean based on symmetric divergence measures, such as the Bhattacharyya divergence, have been considered. The resulting means, although possibly useful in practice, do not satisfy all desirable invariance properties. In this paper we consider geometric means of covariance matrices estimated on high-dimensional time-series, assuming that the data is generated according to an instantaneous mixing model, which is very common in signal processing. We show that in these circumstances we can approximate the Fisher information geometric mean by employing an efficient AJD algorithm. Our approximation is in general much closer to the Fisher information geometric mean as compared to its competitors and verifies many invariance properties. Furthermore, convergence is guaranteed, the computational complexity is low and the convergence rate is quadratic. The accuracy of this new geometric mean approximation is demonstrated by means of simulations.

Multiclass brain-computer interface classification by Riemannian geometry.

This paper presents a new classification framework for brain-computer interface (BCI) based on motor imagery. This framework involves the concept of Riemannian geometry in the manifold of covariance matrices. The main idea is to use spatial covariance matrices as EEG signal descriptors and to rely on Riemannian geometry to directly classify these matrices using the topology of the manifold of symmetric and positive definite (SPD) matrices. This framework allows to extract the spatial information contained in EEG signals without using spatial filtering. Two methods are proposed and compared with a reference method [multiclass Common Spatial Pattern (CSP) and Linear Discriminant Analysis (LDA)] on the multiclass dataset IIa from the BCI Competition IV. The first method, named minimum distance to Riemannian mean (MDRM), is an implementation of the minimum distance to mean (MDM) classification algorithm using Riemannian distance and Riemannian mean. This simple method shows comparable results with the reference method. The second method, named tangent space LDA (TSLDA), maps the covariance matrices onto the Riemannian tangent space where matrices can be vectorized and treated as Euclidean objects. Then, a variable selection procedure is applied in order to decrease dimensionality and a classification by LDA is performed. This latter method outperforms the reference method increasing the mean classification accuracy from 65.1% to 70.2%.