Digital Mobility Measures: A Window into Real-World Severity and Progression of Parkinson's Disease.

BACKGROUND: Real-world monitoring using wearable sensors has enormous potential for assessing disease severity and symptoms among persons with Parkinson's disease (PD). Many distinct features can be extracted, reflecting multiple mobility domains. However, it is unclear which digital measures are related to PD severity and are sensitive to disease progression. OBJECTIVES: The aim was to identify real-world mobility measures that reflect PD severity and show discriminant ability and sensitivity to disease progression, compared to the Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) scale. METHODS: Multicenter real-world continuous (24/7) digital mobility data from 587 persons with PD and 68 matched healthy controls were collected using an accelerometer adhered to the lower back. Machine learning feature selection and regression algorithms evaluated associations of the digital measures using the MDS-UPDRS (I-III). Binary logistic regression assessed discriminatory value using controls, and longitudinal observational data from a subgroup (n = 33) evaluated sensitivity to change over time. RESULTS: Digital measures were only moderately correlated with the MDS-UPDRS (part II-r = 0.60 and parts I and III-r = 0.50). Most associated measures reflected activity quantity and distribution patterns. A model with 14 digital measures accurately distinguished recently diagnosed persons with PD from healthy controls (81.1%, area under the curve: 0.87); digital measures showed larger effect sizes (Cohen's d: [0.19-0.66]), for change over time than any of the MDS-UPDRS parts (Cohen's d: [0.04-0.12]). CONCLUSIONS: Real-world mobility measures are moderately associated with clinical assessments, suggesting that they capture different aspects of motor capacity and function. Digital mobility measures are sensitive to early-stage disease and to disease progression, to a larger degree than conventional clinical assessments, demonstrating their utility, primarily for clinical trials but ultimately also for clinical care. © 2023 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Flexible Temperature Sensor Integration into E-Textiles Using Different Industrial Yarn Fabrication Processes.

Textiles enhanced with thin-film flexible sensors are well-suited for unobtrusive monitoring of skin parameters due to the sensors' high conformability. These sensors can be damaged if they are attached to the surface of the textile, also affecting the textiles' aesthetics and feel. We investigate the effect of embedding flexible temperature sensors within textile yarns, which adds a layer of protection to the sensor. Industrial yarn manufacturing techniques including knit braiding, braiding, and double covering were utilised to identify an appropriate incorporation technique. The thermal time constants recorded by all three sensing yarns was <10 s. Simultaneously, effective sensitivity only decreased by a maximum of 14% compared to the uncovered sensor. This is due to the sensor being positioned within the yarn instead of being in direct contact with the measured surface. These sensor yarns were not affected by bending and produced repeatable measurements. The double covering method was observed to have the least impact on the sensors' performance due to the yarn's smaller dimensions. Finally, a sensing yarn was incorporated in an armband and used to measure changes in skin temperature. The demonstrated textile integration techniques for flexible sensors using industrial yarn manufacturing processes enable large-scale smart textile fabrication.

Thigh-Derived Inertial Sensor Metrics to Assess the Sit-to-Stand and Stand-to-Sit Transitions in the Timed Up and Go (TUG) Task for Quantifying Mobility Impairment in Multiple Sclerosis.

Introduction: Inertial sensors generate objective and sensitive metrics of movement disability that may indicate fall risk in many clinical conditions including multiple sclerosis (MS). The Timed-Up-And-Go (TUG) task is used to assess patient mobility because it incorporates clinically-relevant submovements during standing. Most sensor-based TUG research has focused on the placement of sensors at the spine, hip or ankles; an examination of thigh activity in TUG in multiple sclerosis is wanting. Methods: We used validated sensors (x-IMU by x-io) to derive transparent metrics for the sit-to-stand (SI-ST) transition and the stand-to-sit (ST-SI) transition of TUG, and compared effect sizes for metrics from inertial sensors on the thighs to effect sizes for metrics from a sensor placed at the L3 level of the lumbar spine. Twenty-three healthy volunteers were compared to 17 ambulatory persons with MS (PwMS, HAI ≤ 2). Results: During the SI-ST transition, the metric with the largest effect size comparing healthy volunteers to PwMS was the Area Under the Curve of the thigh angular velocity in the pitch direction-representing both thigh and knee extension; the peak of the spine pitch angular velocity during SI-ST also had a large effect size, as did some temporal measures of duration of SI-ST, although less so. During the ST-SI transition the metric with the largest effect size in PwMS was the peak of the spine angular velocity curve in the roll direction. A regression was performed. Discussion: We propose for PwMS that the diminished peak angular velocity during SI-ST directly represents extensor weakness, while the increased roll during ST-SI represents diminished postural control. Conclusions: During the SI-ST transition of TUG, angular velocities can discriminate between healthy volunteers and ambulatory PwMS better than temporal features. Sensor placement on the thighs provides additional discrimination compared to sensor placement at the lumbar spine.

Deep Convolutional and LSTM Recurrent Neural Networks for Multimodal Wearable Activity Recognition.

Human activity recognition (HAR) tasks have traditionally been solved using engineered features obtained by heuristic processes. Current research suggests that deep convolutional neural networks are suited to automate feature extraction from raw sensor inputs. However, human activities are made of complex sequences of motor movements, and capturing this temporal dynamics is fundamental for successful HAR. Based on the recent success of recurrent neural networks for time series domains, we propose a generic deep framework for activity recognition based on convolutional and LSTM recurrent units, which: (i) is suitable for multimodal wearable sensors; (ii) can perform sensor fusion naturally; (iii) does not require expert knowledge in designing features; and (iv) explicitly models the temporal dynamics of feature activations. We evaluate our framework on two datasets, one of which has been used in a public activity recognition challenge. Our results show that our framework outperforms competing deep non-recurrent networks on the challenge dataset by 4% on average; outperforming some of the previous reported results by up to 9%. Our results show that the framework can be applied to homogeneous sensor modalities, but can also fuse multimodal sensors to improve performance. We characterise key architectural hyperparameters' influence on performance to provide insights about their optimisation.

Ear-worn reference data collection and annotation for multimodal context-aware hearing instruments.

In this work we present a newly developed ear-worn sensing and annotation device to unobtrusively capture head movements in real life situations. It has been designed in the context of developing multimodal hearing instruments (HIs), but is not limited to this application domain. The ear-worn device captures triaxial acceleration, rate of turn and magnetic field and features a one-button-approach for real-time data annotation through the user. The system runtime is over 5 hours at a sampling rate of 128 Hz. In a user study with 21 participants the device was perceived as comfortable and showed a robust hold at the ear. On the example of head acceleration data we perform unsupervised clustering to demonstrate the benefit of head movements for multimodal HIs. We believe the novel technology will help to push the boundaries of HI technology.

Adaptation of hybrid human-computer interaction systems using EEG error-related potentials.

Performance improvement in both humans and artificial systems strongly relies in the ability of recognizing erroneous behavior or decisions. This paper, that builds upon previous studies on EEG error-related signals, presents a hybrid approach for human computer interaction that uses human gestures to send commands to a computer and exploits brain activity to provide implicit feedback about the recognition of such commands. Using a simple computer game as a case study, we show that EEG activity evoked by erroneous gesture recognition can be classified in single trials above random levels. Automatic artifact rejection techniques are used, taking into account that subjects are allowed to move during the experiment. Moreover, we present a simple adaptation mechanism that uses the EEG signal to label newly acquired samples and can be used to re-calibrate the gesture recognition system in a supervised manner. Offline analysis show that, although the achieved EEG decoding accuracy is far from being perfect, these signals convey sufficient information to significantly improve the overall system performance.

Wearable assistant for Parkinson's disease patients with the freezing of gait symptom.

In this paper, we present a wearable assistant for Parkinson's disease (PD) patients with the freezing of gait (FOG) symptom. This wearable system uses on-body acceleration sensors to measure the patients' movements. It automatically detects FOG by analyzing frequency components inherent in these movements. When FOG is detected, the assistant provides a rhythmic auditory signal that stimulates the patient to resume walking. Ten PD patients tested the system while performing several walking tasks in the laboratory. More than 8 h of data were recorded. Eight patients experienced FOG during the study, and 237 FOG events were identified by professional physiotherapists in a post hoc video analysis. Our wearable assistant was able to provide online assistive feedback for PD patients when they experienced FOG. The system detected FOG events online with a sensitivity of 73.1% and a specificity of 81.6%. The majority of patients indicated that the context-aware automatic cueing was beneficial to them. Finally, we characterize the system performance with respect to the walking style, the sensor placement, and the dominant algorithm parameters.