Real-Time Hierarchical Classification of Time Series Data for Locomotion Mode Detection.

OBJECTIVE: Accurate real-time estimation of motion intent is critical for rendering useful assistance using wearable robotic prosthetic and exoskeleton devices during user-initiated motions. We aim to evaluate hierarchical classification as a strategy for real-time locomotion mode recognition for the control of wearable robotic prostheses and exoskeletons during user-initiated motions. METHODS: We collect motion data from 8 subjects using a set of 7 inertial sensors for 16 lower limb locomotion modes of different specificities. A CNN based hierarchical classifier is trained to classify the modes into a specified label hierarchy. We measure the accuracy, stability, behaviour during mode transitions and suitability for real-time inference of the classifier. RESULTS: The method achieves stable classification of locomotion modes using [Formula: see text] of time history data. It achieves average classification accuracy of 94.34% and an average AU(PRC) of 0.773 - comparable to similar classifiers. The method produces more informative classifications at transitions between modes. Less specific classes are classified earlier than more specific classes in the hierarchy. The inference step of the classifier can be executed in less than 2 ms on embedded hardware, indicating suitability for real-time operation. CONCLUSION: Hierarchical classification can achieve accurate detection of locomotion modes and can break up mode transitions into multiple transitions between modes of different specificity. SIGNIFICANCE: Multi-specific hierarchical classification of locomotion modes could lead to smoother, more fine grained control adaptation of wearable robots during locomotion mode transitions.

Schema: metric learning enables interpretable synthesis of heterogeneous single-cell modalities.

A complete understanding of biological processes requires synthesizing information across heterogeneous modalities, such as age, disease status, or gene expression. Technological advances in single-cell profiling have enabled researchers to assay multiple modalities simultaneously. We present Schema, which uses a principled metric learning strategy that identifies informative features in a modality to synthesize disparate modalities into a single coherent interpretation. We use Schema to infer cell types by integrating gene expression and chromatin accessibility data; demonstrate informative data visualizations that synthesize multiple modalities; perform differential gene expression analysis in the context of spatial variability; and estimate evolutionary pressure on peptide sequences.

Assessing single-cell transcriptomic variability through density-preserving data visualization.

Nonlinear data visualization methods, such as t-distributed stochastic neighbor embedding (t-SNE) and uniform manifold approximation and projection (UMAP), summarize the complex transcriptomic landscape of single cells in two dimensions or three dimensions, but they neglect the local density of data points in the original space, often resulting in misleading visualizations where densely populated subsets of cells are given more visual space than warranted by their transcriptional diversity in the dataset. Here we present den-SNE and densMAP, which are density-preserving visualization tools based on t-SNE and UMAP, respectively, and demonstrate their ability to accurately incorporate information about transcriptomic variability into the visual interpretation of single-cell RNA sequencing data. Applied to recently published datasets, our methods reveal significant changes in transcriptomic variability in a range of biological processes, including heterogeneity in transcriptomic variability of immune cells in blood and tumor, human immune cell specialization and the developmental trajectory of Caenorhabditis elegans. Our methods are readily applicable to visualizing high-dimensional data in other scientific domains.

An Omnidirectional Assistive Platform Integrated With Functional Electrical Stimulation for Gait Rehabilitation: A Case Study.

This paper presents a novel omnidirectional platform for gait rehabilitation of people with hemiparesis after stroke. The mobile platform, henceforth the "walker", allows unobstructed pelvic motion during walking, helps the user maintain balance and prevents falls. The system aids mobility actively by combining three types of therapeutic intervention: forward propulsion of the pelvis, controlled body weight support, and functional electrical stimulation (FES) for compensation of deficits in angular motion of the joints. FES is controlled using gait data extracted from a set of inertial measurement units (IMUs) worn by the user. The resulting closed-loop FES system synchronizes stimulation with the gait cycle phases and automatically adapts to the variations in muscle activation caused by changes in residual muscle activity and spasticity. A pilot study was conducted to determine the potential outcomes of the different interventions. One chronic stroke survivor underwent five sessions of gait training, each one involving a total of 30 minutes using the walker and FES system. The patient initially exhibited severe anomalies in joint angle trajectories on both the paretic and the non-paretic side. With training, the patient showed progressive increase in cadence and self-selected gait speed, along with consistent decrease in double-support time. FES helped correct the paretic foot angle during swing phase, and likely was a factor in observed improvements in temporal gait symmetry. Although the experiments showed favorable changes in the paretic trajectories, they also highlighted the need for intervention on the non-paretic side.

Phase-Synchronized Assistive Torque Control for the Correction of Kinematic Anomalies in the Gait Cycle.

Gait anomalies give rise to several clinical problems in stroke survivors, which restrict their functional mobility and have a negative impact on their quality of life. Robotics-aided gait training post-stroke has proven capable of improving patients' functional walking, but so far it has not performed significantly better than conventional therapy. We hypothesize that an exoskeleton-based training program, aimed at correcting deficits in the leg joints' movement, could produce greater improvements in gait function than conventional therapy. As a first step towards testing this hypothesis, we designed an exoskeleton control to correct a typical kinematic deficit post-stroke, namely, reduced knee flexion on the paretic side during swing. The proposed control attempts to minimize this deficit by delivering assistive torque synchronized with the continuous phase of the patient's gait. Nine healthy male participants walked in a unilateral cable-driven exoskeleton while subject to an artificial knee flexion impairment produced by a custom-made knee brace. The experiments employed a treadmill featuring a variable-velocity control to allow self-selected gait speed. The artificial impairment by itself caused a significant reduction in peak flexion angle (p = 0.000129). Exoskeleton assistance compensated most of the knee flexion deficit, yielding no significant difference with unrestricted flexion (p = 0.3393). No significant changes in self-selected gait speed or stride frequency were detected. The proposed control can be expanded to correct motion deficits in other joints at different stages of the gait cycle.

High mobility control of an omnidirectional platform for gait rehabilitation after stroke.

We present a novel control method for an omnidirectional robotic platform for gait training. This mobile platform or "walker" provides trunk support and allows unrestricted motion of the pelvis simultaneously. In addition to helping the user maintain balance and preventing falls, the walker combines two types of therapeutic intervention: forward propulsion of the trunk and partial body weight support (BWS). The core of the walker's control is an admittance controller that maximizes the platform's horizontal mobility by optimizing the virtual mass of the admittance model. Said mass represents the best tradeoff between a low-frequency oscillation mode that becomes more damped as the virtual mass decreases, and a high-frequency mode that becomes less damped simultaneously and hence could destabilize the system. Forward propulsion of the trunk is aided by a horizontal force that is modulated with the patient's gait speed and turning rate to ensure easy adaptation. BWS is provided by a second, independent admittance controller that generates a spring-like upward force. In an initial study, a stroke patient was able to walk stably in the platform, as evidenced by the absence of oscillations associated with an excessively low virtual mass. A progressive increase in the patient's self-selected speed, along with greater uniformity in the instantaneous velocity, suggest that forward propulsion was effective in compensating the patient's own propulsion deficit.