Physiological sensor data cleaning with autoencoders.

Objective.Physiological sensor data (e.g. photoplethysmograph) is important for remotely monitoring patients' vital signals, but is often affected by measurement noise. Existing feature-based models for signal cleaning can be limited as they might not capture the full signal characteristics.Approach.In this work we present a deep learning framework for sensor signal cleaning based on dilated convolutions which capture the coarse- and fine-grained structure in order to classify whether a signal is noisy or clean. However, since obtaining annotated physiological data is costly and time-consuming we propose an autoencoder-based semi-supervised model which is able to learn a representation of the sensor signal characteristics, also adding an element of interpretability.Main results.Our proposed models are over 8% more accurate than existing feature-based approaches with half the false positive/negative rates. Finally, we show that with careful tuning (that can be improved further), the semi-supervised model outperforms supervised approaches suggesting that incorporating the large amounts of available unlabeled data can be advantageous for achieving high accuracy (over 90%) and minimizing the false positive/negative rates.Significance.Our approach enables us to reliably separate clean from noisy physiological sensor signal that can pave the development of reliable features and eventually support decisions regarding drug efficacy in clinical trials.

Assessment of Upper Extremity Function in Multiple Sclerosis: Feasibility of a Digital Pinching Test.

BACKGROUND: The development of touchscreen-based assessments of upper extremity function could benefit people with multiple sclerosis (MS) by allowing convenient, quantitative assessment of their condition. The Pinching Test forms a part of the Floodlight smartphone app (F. Hoffmann-La Roche Ltd, Basel, Switzerland) for people with MS and was designed to capture upper extremity function. OBJECTIVE: This study aimed to evaluate the Pinching Test as a tool for remotely assessing upper extremity function in people with MS. METHODS: Using data from the 24-week, prospective feasibility study investigating the Floodlight Proof-of-Concept app for remotely assessing MS, we examined 13 pinching, 11 inertial measurement unit (IMU)-based, and 13 fatigability features of the Pinching Test. We assessed the test-retest reliability using intraclass correlation coefficients [second model, first type; ICC(2,1)], age- and sex-adjusted cross-sectional Spearman rank correlation, and known-groups validity (data aggregation: median [all features], SD [fatigability features]). RESULTS: We evaluated data from 67 people with MS (mean Expanded Disability Status Scale [EDSS]: 2.4 [SD 1.4]) and 18 healthy controls. In this cohort of early MS, pinching features were reliable [ICC(2,1)=0.54-0.81]; correlated with standard clinical assessments, including the Nine-Hole Peg Test (9HPT) (|r|=0.26-0.54; 10/13 features), EDSS (|r|=0.25-0.36; 7/13 features), and the arm items of the 29-item Multiple Sclerosis Impact Scale (MSIS-29) (|r|=0.31-0.52; 7/13 features); and differentiated people with MS-Normal from people with MS-Abnormal (area under the curve: 0.68-0.78; 8/13 features). IMU-based features showed similar test-retest reliability [ICC(2,1)=0.47-0.84] but showed little correlations with standard clinical assessments. In contrast, fatigability features (SD aggregation) correlated with 9HPT time (|r|=0.26-0.61; 10/13 features), EDSS (|r|=0.26-0.41; 8/13 features), and MSIS-29 arm items (|r|=0.32-0.46; 7/13 features). CONCLUSIONS: The Pinching Test provides a remote, objective, and granular assessment of upper extremity function in people with MS that can potentially complement standard clinical evaluation. Future studies will validate it in more advanced MS. TRIAL REGISTRATION: ClinicalTrials.gov NCT02952911; https://clinicaltrials.gov/study/NCT02952911.

Digital measures of respiratory and upper limb function in spinal muscular atrophy: design, feasibility, reliability, and preliminary validity of a smartphone sensor-based assessment suite.

Spinal muscular atrophy (SMA) is characterized by progressive muscle weakness and paralysis. Motor function is monitored in the clinical setting using assessments including the 32-item Motor Function Measure (MFM-32), but changes in disease severity between clinical visits may be missed. Digital health technologies may assist evaluation of disease severity by bridging gaps between clinical visits. We developed a smartphone sensor-based assessment suite, comprising nine tasks, to assess motor and muscle function in people with SMA. We used data from the risdiplam phase 2 JEWELFISH trial to assess the test-retest reliability and convergent validity of each task. In the first 6 weeks, 116 eligible participants completed assessments on a median of 6.3 days per week. Eight of the nine tasks demonstrated good or excellent test-retest reliability (intraclass correlation coefficients >0.75 and >0.9, respectively). Seven tasks showed a significant association (P < 0.05) with related clinical measures of motor function (individual items from the MFM-32 or Revised Upper Limb Module scales) and seven showed significant association (P < 0.05) with disease severity measured using the MFM-32 total score. This cross-sectional study supports the feasibility, reliability, and validity of using smartphone-based digital assessments to measure function in people living with SMA.

Automatic speaker diarization for natural conversation analysis in autism clinical trials.

Challenges in social communication is one of the core symptom domains in autism spectrum disorder (ASD). Novel therapies are under development to help individuals with these challenges, however the ability to show a benefit is dependent on a sensitive and reliable measure of treatment effect. Currently, measuring these deficits requires the use of time-consuming and subjective techniques. Objective measures extracted from natural conversations could be more ecologically relevant, and administered more frequently-perhaps giving them added sensitivity to change. While several studies have used automated analysis methods to study autistic speech, they require manual transcriptions. In order to bypass this time-consuming process, an automated speaker diarization algorithm must first be applied. In this paper, we are testing whether a speaker diarization algorithm can be applied to natural conversations between autistic individuals and their conversational partner in a natural setting at home over the course of a clinical trial. We calculated the average duration that a participant would speak for within their turn. We found a significant correlation between this feature and the Vineland Adaptive Behaviour Scales (VABS) expressive communication score (r = 0.51, p = 7 × 10-5). Our results show that natural conversations can be used to obtain measures of talkativeness, and that this measure can be derived automatically, thus showing the promise of objectively evaluating communication challenges in ASD.

Personalized Longitudinal Assessment of Multiple Sclerosis Using Smartphones.

Personalized longitudinal disease assessment is central to quickly diagnosing, appropriately managing, and optimally adapting the therapeutic strategy of multiple sclerosis (MS). It is also important for identifying idiosyncratic subject-specific disease profiles. Here, we design a novel longitudinal model to map individual disease trajectories in an automated way using smartphone sensor data that may contain missing values. First, we collect digital measurements related to gait and balance, and upper extremity functions using sensor-based assessments administered on a smartphone. Next, we treat missing data via imputation. We then discover potential markers of MS by employing a generalized estimation equation. Subsequently, parameters learned from multiple training datasets are ensembled to form a simple, unified longitudinal predictive model to forecast MS over time in previously unseen people with MS. To mitigate potential underestimation for individuals with severe disease scores, the final model incorporates additional subject-specific fine-tuning using data from the first day. The results show that the proposed model is promising to achieve personalized longitudinal MS assessment; they also suggest that features related to gait and balance as well as upper extremity function, remotely collected from sensor-based assessments, may be useful digital markers for predicting MS over time.

Preliminary validity of the Draw a Shape Test for upper extremity assessment in multiple sclerosis.

OBJECTIVE: To validate the smartphone sensor-based Draw a Shape Test - a part of the Floodlight Proof-of-Concept app for remotely assessing multiple sclerosis-related upper extremity impairment by tracing six different shapes. METHODS: People with multiple sclerosis, classified functionally normal/abnormal via their Nine-Hole Peg Test time, and healthy controls participated in a 24-week, nonrandomized study. Spatial (trace accuracy), temporal (mean and variability in linear, angular, and radial drawing velocities, and dwell time ratio), and spatiotemporal features (trace celerity) were cross-sectionally analyzed for correlation with standard clinical and brain magnetic resonance imaging (normalized brain volume and total lesion volume) disease burden measures, and for capacity to differentiate people with multiple sclerosis from healthy controls. RESULTS: Data from 69 people with multiple sclerosis and 18 healthy controls were analyzed. Trace accuracy (all shapes), linear velocity variability (circle, figure-of-8, spiral shapes), and radial velocity variability (spiral shape) had a mostly fair/moderate-to-good correlation (|r| = 0.14-0.66) with all disease burden measures. Trace celerity also had mostly fair/moderate-to-good correlation (|r| = 0.18-0.41) with Nine-Hole Peg Test performance, cerebellar functional system score, and brain magnetic resonance imaging. Furthermore, partial correlation analysis related these results to motor impairment. People with multiple sclerosis showed greater drawing velocity variability, though slower mean velocity, than healthy controls. Linear velocity (spiral shape) and angular velocity (circle shape) potentially differentiate functionally normal people with multiple sclerosis from healthy controls. INTERPRETATION: The Draw a Shape Test objectively assesses upper extremity impairment and correlates with all disease burden measures, thus aiding multiple sclerosis-related upper extremity impairment characterization.

Neural correlates of digital measures shown by structural MRI: a post-hoc analysis of a smartphone-based remote assessment feasibility study in multiple sclerosis.

BACKGROUND: A study was undertaken to evaluate remote monitoring via smartphone sensor-based tests in people with multiple sclerosis (PwMS). This analysis aimed to explore regional neural correlates of digital measures derived from these tests. METHODS: In a 24-week, non-randomized, interventional, feasibility study (NCT02952911), sensor-based tests on the Floodlight Proof-of-Concept app were used to assess cognition (smartphone-based electronic Symbol Digit Modalities Test), upper extremity function (Draw a Shape Test, Pinching Test), and gait and balance (Static Balance Test, Two-Minute Walk Test, U-Turn Test). In this post-hoc analysis, digital measures and standard clinical measures (e.g., Nine-Hole Peg Test [9HPT]) were correlated against regional structural magnetic resonance imaging outcomes. Seventy-six PwMS aged 18-55 years with an Expanded Disability Status Scale score of 0.0-5.5 were enrolled from two different sites (USA and Spain). Sixty-two PwMS were included in this analysis. RESULTS: Worse performance on digital and clinical measures was associated with smaller regional brain volumes and larger ventricular volumes. Whereas digital and clinical measures had many neural correlates in common (e.g., putamen, globus pallidus, caudate nucleus, lateral occipital cortex), some were observed only for digital measures. For example, Draw a Shape Test and Pinching Test measures, but not 9HPT score, correlated with volume of the hippocampus (r = 0.37 [drawing accuracy over time on the Draw a Shape Test]/ - 0.45 [touching asynchrony on the Pinching Test]), thalamus (r = 0.38/ - 0.41), and pons (r = 0.35/ - 0.35). CONCLUSIONS: Multiple neural correlates were identified for the digital measures in a cohort of people with early MS. Digital measures showed associations with brain regions that clinical measures were unable to demonstrate, thus providing potential novel information on functional ability compared with standard clinical assessments.

Longitudinal Trend Monitoring of Multiple Sclerosis Ambulation Using Smartphones.

Goal: Smartphone and wearable devices may act as powerful tools to remotely monitor physical function in people with neurodegenerative and autoimmune diseases from out-of-clinic environments. Detection of progression onset or worsening of symptoms is especially important in people living with multiple sclerosis (PwMS) in order to enable optimally adapted therapeutic strategies. MS symptoms typically follow subtle and fluctuating disease courses, patient-to-patient, and over time. Current in-clinic assessments are often too infrequently administered to reflect longitudinal changes in MS impairment that impact daily life. This work, therefore, explores how smartphones can administer daily two-minute walking assessments to monitor PwMS physical function at home. Methods: Remotely collected smartphone inertial sensor data was transformed through state-of-the-art Deep Convolutional Neural Networks, to estimate a participant's daily ambulatory-related disease severity, longitudinally over a 24-week study. Results: This study demonstrated that smartphone-based ambulatory severity outcomes could accurately estimate MS level of disability, as measured by the EDSS score ([Formula: see text]: 0.56,[Formula: see text]0.001). Furthermore, longitudinal severity outcomes were shown to accurately reflect individual participants' level of disability over the study duration. Conclusion: Smartphone-based assessments, that can be performed by patients from their home environments, could greatly augment standard in-clinic outcomes for neurodegenerative diseases. The ability to understand the impact of disease on daily-life between clinical visits, through objective digital outcomes, paves the way forward to better measure and identify signs of disease progression that may be occurring out-of-clinic, to monitor how different patients respond to various treatments, and to ultimately enable the development of better, and more personalised care.

Practice effects in performance outcome measures in patients living with neurologic disorders - A systematic review.

Background: In this systematic review we sought to characterize practice effects on traditional in-clinic or digital performance outcome measures commonly used in one of four neurologic disease areas (multiple sclerosis; Huntington's disease; Parkinson's disease; and Alzheimer's disease, mild cognitive impairment and other forms of dementia), describe mitigation strategies to minimize their impact on data interpretation and identify gaps to be addressed in future work. Methods: Fifty-eight original articles (49 from Embase and an additional 4 from PubMed and 5 from additional sources; cut-off date January 13, 2021) describing practice effects or their mitigation strategies were included. Results: Practice effects observed in healthy volunteers do not always translate to patients living with neurologic disorders. Mitigation strategies include reliable changes indices that account for practice effects or a run-in period. While the former requires data from a reference sample showing similar practice effects, the latter requires a sufficient number of tests in the run-in period to reach steady-state performance. However, many studies only included 2 or 3 test administrations, which is insufficient to define the number of tests needed in a run-in period. Discussion: Several gaps have been identified. In particular the assessment of practice effects on an individual patient level as well as the temporal dynamics of practice effects are largely unaddressed. Here, digital tests, which allow much higher testing frequency over prolonged periods of time, can be used in future work to gain a deeper understanding of practice effects and to develop new metrics for assessing and accounting for practice effects in clinical research and clinical trials.

Trial of Prasinezumab in Early-Stage Parkinson's Disease.

BACKGROUND: Aggregated α-synuclein plays an important role in the pathogenesis of Parkinson's disease. The monoclonal antibody prasinezumab, directed at aggregated α-synuclein, is being studied for its effect on Parkinson's disease. METHODS: In this phase 2 trial, we randomly assigned participants with early-stage Parkinson's disease in a 1:1:1 ratio to receive intravenous placebo or prasinezumab at a dose of 1500 mg or 4500 mg every 4 weeks for 52 weeks. The primary end point was the change from baseline to week 52 in the sum of scores on parts I, II, and III of the Movement Disorder Society-sponsored revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS; range, 0 to 236, with higher scores indicating greater impairment). Secondary end points included the dopamine transporter levels in the putamen of the hemisphere ipsilateral to the clinically more affected side of the body, as measured by 123I-ioflupane single-photon-emission computed tomography (SPECT). RESULTS: A total of 316 participants were enrolled; 105 were assigned to receive placebo, 105 to receive 1500 mg of prasinezumab, and 106 to receive 4500 mg of prasinezumab. The baseline mean MDS-UPDRS scores were 32.0 in the placebo group, 31.5 in the 1500-mg group, and 30.8 in the 4500-mg group, and mean (±SE) changes from baseline to 52 weeks were 9.4±1.2 in the placebo group, 7.4±1.2 in the 1500-mg group (difference vs. placebo, -2.0; 80% confidence interval [CI], -4.2 to 0.2; P = 0.24), and 8.8±1.2 in the 4500-mg group (difference vs. placebo, -0.6; 80% CI, -2.8 to 1.6; P = 0.72). There was no substantial difference between the active-treatment groups and the placebo group in dopamine transporter levels on SPECT. The results for most clinical secondary end points were similar in the active-treatment groups and the placebo group. Serious adverse events occurred in 6.7% of the participants in the 1500-mg group and in 7.5% of those in the 4500-mg group; infusion reactions occurred in 19.0% and 34.0%, respectively. CONCLUSIONS: Prasinezumab therapy had no meaningful effect on global or imaging measures of Parkinson's disease progression as compared with placebo and was associated with infusion reactions. (Funded by F. Hoffmann-La Roche and Prothena Biosciences; PASADENA ClinicalTrials.gov number, NCT03100149.).

Reliability and validity of the Roche PD Mobile Application for remote monitoring of early Parkinson's disease.

Digital health technologies enable remote and therefore frequent measurement of motor signs, potentially providing reliable and valid estimates of motor sign severity and progression in Parkinson's disease (PD). The Roche PD Mobile Application v2 was developed to measure bradykinesia, bradyphrenia and speech, tremor, gait and balance. It comprises 10 smartphone active tests (with ½ tests administered daily), as well as daily passive monitoring via a smartphone and smartwatch. It was studied in 316 early-stage PD participants who performed daily active tests at home then carried a smartphone and wore a smartwatch throughout the day for passive monitoring (study NCT03100149). Here, we report baseline data. Adherence was excellent (96.29%). All pre-specified sensor features exhibited good-to-excellent test-retest reliability (median intraclass correlation coefficient = 0.9), and correlated with corresponding Movement Disorder Society-Unified Parkinson's Disease Rating Scale items (rho: 0.12-0.71). These findings demonstrate the preliminary reliability and validity of remote at-home quantification of motor sign severity with the Roche PD Mobile Application v2 in individuals with early PD.

Multiscale topology characterizes dynamic tumor vascular networks.

Advances in imaging techniques enable high-resolution three-dimensional (3D) visualization of vascular networks over time and reveal abnormal structural features such as twists and loops, and their quantification is an active area of research. Here, we showcase how topological data analysis, the mathematical field that studies the "shape" of data, can characterize the geometric, spatial, and temporal organization of vascular networks. We propose two topological lenses to study vasculature, which capture inherent multiscale features and vessel connectivity, and surpass the single-scale analysis of existing methods. We analyze images collected using intravital and ultramicroscopy modalities and quantify spatiotemporal variation of twists, loops, and avascular regions (voids) in 3D vascular networks. This topological approach validates and quantifies known qualitative trends such as dynamic changes in tortuosity and loops in response to antibodies that modulate vessel sprouting; furthermore, it quantifies the effect of radiotherapy on vessel architecture.

A Remote Digital Monitoring Platform to Assess Cognitive and Motor Symptoms in Huntington Disease: Cross-sectional Validation Study.

BACKGROUND: Remote monitoring of Huntington disease (HD) signs and symptoms using digital technologies may enhance early clinical diagnosis and tracking of disease progression, guide treatment decisions, and monitor response to disease-modifying agents. Several recent studies in neurodegenerative diseases have demonstrated the feasibility of digital symptom monitoring. OBJECTIVE: The aim of this study was to evaluate a novel smartwatch- and smartphone-based digital monitoring platform to remotely monitor signs and symptoms of HD. METHODS: This analysis aimed to determine the feasibility and reliability of the Roche HD Digital Monitoring Platform over a 4-week period and cross-sectional validity over a 2-week interval. Key criteria assessed were feasibility, evaluated by adherence and quality control failure rates; test-retest reliability; known-groups validity; and convergent validity of sensor-based measures with existing clinical measures. Data from 3 studies were used: the predrug screening phase of an open-label extension study evaluating tominersen (NCT03342053) and 2 untreated cohorts-the HD Natural History Study (NCT03664804) and the Digital-HD study. Across these studies, controls (n=20) and individuals with premanifest (n=20) or manifest (n=179) HD completed 6 motor and 2 cognitive tests at home and in the clinic. RESULTS: Participants in the open-label extension study, the HD Natural History Study, and the Digital-HD study completed 89.95% (1164/1294), 72.01% (2025/2812), and 68.98% (1454/2108) of the active tests, respectively. All sensor-based features showed good to excellent test-retest reliability (intraclass correlation coefficient 0.89-0.98) and generally low quality control failure rates. Good overall convergent validity of sensor-derived features to Unified HD Rating Scale outcomes and good overall known-groups validity among controls, premanifest, and manifest participants were observed. Among participants with manifest HD, the digital cognitive tests demonstrated the strongest correlations with analogous in-clinic tests (Pearson correlation coefficient 0.79-0.90). CONCLUSIONS: These results show the potential of the HD Digital Monitoring Platform to provide reliable, valid, continuous remote monitoring of HD symptoms, facilitating the evaluation of novel treatments and enhanced clinical monitoring and care for individuals with HD.

Measuring Health-Related Quality of Life With Multimodal Data: Viewpoint.

The ability to objectively measure aspects of performance and behavior is a fundamental pillar of digital health, enabling digital wellness products, decentralized trial concepts, evidence generation, digital therapeutics, and more. Emerging multimodal technologies capable of measuring several modalities simultaneously and efforts to integrate inputs across several sources are further expanding the limits of what digital measures can assess. Experts from the field of digital health were convened as part of a multi-stakeholder workshop to examine the progress of multimodal digital measures in two key areas: detection of disease and the measurement of meaningful aspects of health relevant to the quality of life. Here we present a meeting report, summarizing key discussion points, relevant literature, and finally a vision for the immediate future, including how multimodal measures can provide value to stakeholders across drug development and care delivery, as well as three key areas where headway will need to be made if we are to continue to build on the encouraging progress so far: collaboration and data sharing, removal of barriers to data integration, and alignment around robust modular evaluation of new measurement capabilities.

A smartphone sensor-based digital outcome assessment of multiple sclerosis.

BACKGROUND: Sensor-based monitoring tools fill a critical gap in multiple sclerosis (MS) research and clinical care. OBJECTIVE: The aim of this study is to assess performance characteristics of the Floodlight Proof-of-Concept (PoC) app. METHODS: In a 24-week study (clinicaltrials.gov: NCT02952911), smartphone-based active tests and passive monitoring assessed cognition (electronic Symbol Digit Modalities Test), upper extremity function (Pinching Test, Draw a Shape Test), and gait and balance (Static Balance Test, U-Turn Test, Walk Test, Passive Monitoring). Intraclass correlation coefficients (ICCs) and age- or sex-adjusted Spearman's rank correlation determined test-retest reliability and correlations with clinical and magnetic resonance imaging (MRI) outcome measures, respectively. RESULTS: Seventy-six people with MS (PwMS) and 25 healthy controls were enrolled. In PwMS, ICCs were moderate-to-good (ICC(2,1) = 0.61-0.85) across tests. Correlations with domain-specific standard clinical disability measures were significant for all tests in the cognitive (r = 0.82, p < 0.001), upper extremity function (|r|= 0.40-0.64, all p < 0.001), and gait and balance domains (r = -0.25 to -0.52, all p < 0.05; except for Static Balance Test: r = -0.20, p > 0.05). Most tests also correlated with Expanded Disability Status Scale, 29-item Multiple Sclerosis Impact Scale items or subscales, and/or normalized brain volume. CONCLUSION: The Floodlight PoC app captures reliable and clinically relevant measures of functional impairment in MS, supporting its potential use in clinical research and practice.

Self-Report versus Clinician Examination in Early Parkinson's Disease.

BACKGROUND: Evaluating the discrepancies between patient-reported measures and clinician examination has implications for formulating individual treatment regimens. OBJECTIVE: This study investigated the association between health outcomes and level of self-reported motor-related function impairment relative to clinician-examined motor signs. METHODS: Recently diagnosed PD patients were evaluated using the Parkinson's Progression Marker Initiative (PPMI, N = 420) and the PASADENA phase II clinical trial (N = 316). We calculated the average normalized difference between each participant's part II and III MDS-UPDRS (Movement Disorder Society Unified Parkinson's Disease Rating Scale) scores. Individuals with score differences <25th or >75th percentiles were labeled as low- and high-self-reporters, respectively (those between ranges were labeled intermediate-self-reporters). We compared a wide range of clinical/biomarker readouts among these three groups, using Kruskal-Wallis nonparametric and Pearson's χ2 tests. Spearman's correlations were tested for associations between MDS-UPDRS subscales. RESULTS: In both cohorts, high-self-reporters reported the largest impairment/symptom experience for most motor and nonmotor patient-reported variables. By contrast, these high-self-reporters were similar to or less impaired on clinician-examined and biomarker measures. Patient-reported nonmotor symptoms on MDS-UPDRS part IB showed the strongest positive correlation with self-reported motor-related impairment (PPMI rs  = 0.54, PASADENA rs  = 0.52). This correlation was numerically stronger than the part II and clinician-examined MDS-UPDRS part III correlation (PPMI rs  = 0.38, PASADENA rs  = 0.28). CONCLUSION: Self-reported motor-related impairments reflect not only motor signs/symptoms but also other self-reported nonmotor measures. This may indicate (1) a direct impact of nonmotor symptoms on motor-related functioning and/or (2) the existence of general response tendencies in how patients self-rate symptoms. Our findings suggest further investigation into the suitability of MDS-UPDRS II to assess motor-related impairments. © 2021 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Simulating the impact of noise on gait features extracted from smartphone sensor-data for the remote assessment of movement disorders.

Signs and symptoms of movement disorders can be remotely measured at home through sensor-based assessment of gait. However, sensor noise may impact the robustness of such assessments, in particular in a Bring-Your-Own-Device setting where the quality of sensors might vary. Here, we propose a framework to study the impact of inertial measurement unit noise on sensor-based gait features. This framework includes synthesizing realistic acceleration signals from the lower back during a gait cycle in OpenSim, estimating the magnitude of sensor noise from five smartphone models, perturbing the synthesized acceleration signal with the estimated noise in a Monte Carlo simulation, and computing gait features. In addition, we show that realistic levels of sensor noise have only a negligible impact on step power, a measure of gait.

Interpretable deep learning for the remote characterisation of ambulation in multiple sclerosis using smartphones.

The emergence of digital technologies such as smartphones in healthcare applications have demonstrated the possibility of developing rich, continuous, and objective measures of multiple sclerosis (MS) disability that can be administered remotely and out-of-clinic. Deep Convolutional Neural Networks (DCNN) may capture a richer representation of healthy and MS-related ambulatory characteristics from the raw smartphone-based inertial sensor data than standard feature-based methodologies. To overcome the typical limitations associated with remotely generated health data, such as low subject numbers, sparsity, and heterogeneous data, a transfer learning (TL) model from similar large open-source datasets was proposed. Our TL framework leveraged the ambulatory information learned on human activity recognition (HAR) tasks collected from wearable smartphone sensor data. It was demonstrated that fine-tuning TL DCNN HAR models towards MS disease recognition tasks outperformed previous Support Vector Machine (SVM) feature-based methods, as well as DCNN models trained end-to-end, by upwards of 8-15%. A lack of transparency of "black-box" deep networks remains one of the largest stumbling blocks to the wider acceptance of deep learning for clinical applications. Ensuing work therefore aimed to visualise DCNN decisions attributed by relevance heatmaps using Layer-Wise Relevance Propagation (LRP). Through the LRP framework, the patterns captured from smartphone-based inertial sensor data that were reflective of those who are healthy versus people with MS (PwMS) could begin to be established and understood. Interpretations suggested that cadence-based measures, gait speed, and ambulation-related signal perturbations were distinct characteristics that distinguished MS disability from healthy participants. Robust and interpretable outcomes, generated from high-frequency out-of-clinic assessments, could greatly augment the current in-clinic assessment picture for PwMS, to inform better disease management techniques, and enable the development of better therapeutic interventions.

Beacon-Based Remote Measurement of Social Behavior in ASD Clinical Trials: A Technical Feasibility Assessment.

In this work, we propose a Bluetooth low energy (BLE) beacon-based algorithm to enable remote measurement of the social behavior of the participants of an observational Autism Spectrum Disorder (ASD) clinical trial (NCT03611075). We have developed a mobile application for a smartphone and a smartwatch to collect beacon signals from BLE beacon sensors as well as to store information about the participants' household rooms. Our goal is to collect beacon information about the time the participants spent in different rooms of their household to infer sociability information. We applied the same technology and setup in an internal experiment with healthy volunteers to evaluate the accuracy of the proposed algorithm in 10 different home setups, and we observed an average accuracy of 97.2%. Moreover, we show that it is feasible for the clinical study participants/caregivers to set up the BLE beacon sensors in their homes without any technical help, with 96% of them setting up the technology on the first day of data collection. Next, we present results from one-week location data from study participants collected through the proposed technology. Finally, we provide a list of good practice guidelines for optimally applying beacon technology for indoor location monitoring. The proposed algorithm enables us to estimate time spent in different rooms of a household that can pave the development of objective sociability features and eventually support decisions regarding drug efficacy in ASD.

Smartphone- and Smartwatch-Based Remote Characterisation of Ambulation in Multiple Sclerosis During the Two-Minute Walk Test.

Leveraging consumer technology such as smartphone and smartwatch devices to objectively assess people with multiple sclerosis (PwMS) remotely could capture unique aspects of disease progression. This study explores the feasibility of assessing PwMS and Healthy Control's (HC) physical function by characterising gait-related features, which can be modelled using machine learning (ML) techniques to correctly distinguish subgroups of PwMS from healthy controls. A total of 97 subjects (24 HC subjects, 52 mildly disabled (PwMSmild, EDSS [0-3]) and 21 moderately disabled (PwMSmod, EDSS [3.5-5.5]) contributed data which was recorded from a Two-Minute Walk Test (2MWT) performed out-of-clinic and daily over a 24-week period. Signal-based features relating to movement were extracted from sensors in smartphone and smartwatch devices. A large number of features (n = 156) showed fair-to-strong (R 0.3) correlations with clinical outcomes. LASSO feature selection was applied to select and rank subsets of features used for dichotomous classification between subject groups, which were compared using Logistic Regression (LR), Support Vector Machines (SVM) and Random Forest (RF) models. Classifications of subject types were compared using data obtained from smartphone, smartwatch and the fusion of features from both devices. Models built on smartphone features alone achieved the highest classification performance, indicating that accurate and remote measurement of the ambulatory characteristics of HC and PwMS can be achieved with only one device. It was observed however that smartphone-based performance was affected by inconsistent placement location (running belt versus pocket). Results show that PwMSmod could be distinguished from HC subjects (Acc. 82.2 ± 2.9%, Sen. 80.1 ± 3.9%, Spec. 87.2 ± 4.2%, F 1 84.3 ± 3.8), and PwMSmild (Acc. 82.3 ± 1.9%, Sen. 71.6 ± 4.2%, Spec. 87.0 ± 3.2%, F 1 75.1 ± 2.2) using an SVM classifier with a Radial Basis Function (RBF). PwMSmild were shown to exhibit HC-like behaviour and were thus less distinguishable from HC (Acc. 66.4 ± 4.5%, Sen. 67.5 ± 5.7%, Spec. 60.3 ± 6.7%, F 1 58.6 ± 5.8). Finally, it was observed that subjects in this study demonstrated low intra- and high inter-subject variability which was representative of subject-specific gait characteristics.