Digital Biomarkers最新文献

Introduction: Changes in speech can act as biomarkers of cognitive decline in Alzheimer's disease (AD). While shorter speech samples would promote data collection and analysis, the minimum length of informative speech samples remains debated. This study aims to provide insight into the effect of sample length in analyzing longitudinal recordings of spontaneous speech in AD by comparing the original random length, 5- and 1-minute-long samples. We hope to understand whether capping the audio improves the accuracy of the analysis, and whether an extra 4 min conveys necessary information.

Methods: 110 spontaneous speech samples were collected from decades of Youtube videos of 17 public figures, 9 of whom eventually developed AD. 456 language features were extracted and their text-length-sensitivity, comparability, and ability to capture change over time were analyzed across three different sample lengths.

Results: Capped audio files had advantages over the random length ones. While most extracted features were statistically comparable or highly correlated across the datasets, potential effects of sample length should be acknowledged for some features. The 5-min dataset presented the highest reliability in tracking the evolution of the disease, suggesting that the 4 extra minutes do convey informative data.

Conclusion: Sample length seems to play an important role in extracting the language feature values from speech and tracking disease progress over time. We highlight the importance of further research into optimal sample length and standardization of methods when studying speech in AD.

Introduction: Head impulse, nystagmus, and test of skew (HINTS) is more accurate for the early diagnosis of occipital fossa stroke than magnetic resonance imaging. However, the head impulse test (HIT) is relatively challenging to perform, as it is subjective. Herein, we developed a prototype video HIT (vHIT) system using an iPhone (Apple, Cupertino, CA, USA) that is compact, easy to operate, and analyzable by our iPhone application.

Methods: The iPhone-vHIT and a vHIT using EyeSeeCam (Interacoustics, Eden Prairie, NM, USA) were performed on a healthy man in his 30s and on a patient with vestibular neuritis who visited the Mejiro University Ear Institute Clinic. For the iPhone-vHIT, eye movements were detected by analyzing high-speed videos captured using an iPhone camera, and head movements were followed using an iPhone gyro sensor. An iPhone fixation brace was used to capture the video without any blurring.

Results: The iPhone-vHIT system obtained vHIT waveforms similar to those of the EyeSeeCam-vHIT system in the healthy man and the patient with vestibular neuritis. The iPhone-vHIT system effectively detected the reduced vestibulo-ocular reflex gain in patients with vestibular neuritis. The iPhone-vHIT system at 120 frames per second was less sensitive to catch-up saccades than the EyeSeeCam.

Conclusion: vHIT systems using a smartphone have been reported but are currently unavailable. At present, the iPhone-vHIT application in this study is the only available smartphone-based vHIT system for screening of peripheral vestibular dysfunction. We believe that the prototype iPhone-vHIT with a high-speed camera will be clinically used to perform the vHIT, even though it only examines the lateral semicircular canal.

Introduction: Exposure to light fundamentally influences human physiology and behaviour by synchronising our biological clock to the external light-dark cycle and controlling melatonin production. In addition to well-controlled laboratory studies, more naturalistic approaches to examining these "non-visual" effects of light have been developed in recent years. As naturalistic light exposure is quite unlike well-controlled stimulus conditions in the laboratory, it is critical to measure light exposure in a person-referenced way, the "spectral diet." To this end, light loggers have been developed to capture personalised light exposure. As an alternative to light sensors integrated into wrist-worn actimeters, pendants, or brooch-based light loggers, a recently developed wearable light logger laterally attached to spectacle frames enables the measurement of biologically relevant quantities in the corneal plane.

Methods: Here, we examine the usability and acceptability of using the light logger in an undergraduate student sample (n = 18, mean±1SD: 20.1 ± 1.7 years; 9 female; Oxford, UK) in real-world conditions during a 24-h measurement period. We probed the acceptability of the light logger using rating questionnaires and open-ended questions.

Results: Our quantitative results show a modest acceptability of the light logger. A thematic analysis of the open-ended questions reveals that the form factor of the device, in particular, size, weight, and stability, and reactions from other people to the wearer of the light logger, were commonly mentioned aspects.

Conclusion: In sum, the results indicate the miniaturisation of light loggers and "invisible" integration into extant everyday objects as key areas for future technological development, facilitating the availability of light exposure data for developing personalised intervention strategies in both research, clinical and consumer contexts.

Background: Innovative Medicines Initiative (IMI) consortium IDEA-FAST is developing novel digital measures of fatigue, sleep quality, and impact of sleep disturbances for neurodegenerative diseases and immune-mediated inflammatory diseases. In 2022, the consortium met with the European Medicines Agency (EMA) to receive advice on its plans for regulatory qualification of the measures. This viewpoint reviews the IDEA-FAST perspective on developing digital measures for multiple diseases and the advice provided by the EMA.

Summary: The EMA considered a cross-disease measure an interesting and arguably feasible concept. Developers should account for the need for a strong rationale that the clinical features to be measured are similar across diseases. In addition, they may expect increased complexity of study design, challenges when managing differences within and between disease populations, and the need for validation in both heterogeneous and homogeneous populations.

Key messages: EMA highlighted the challenges teams may encounter when developing a cross-disease measure, though benefits potentially include reduced resources for the technology developer and health authority, faster access to innovation across different therapeutic fields, and feasibility of cross-disease comparisons. The insights included here can be used by project teams to guide them in the development of cross-disease digital measures intended for regulatory qualification.

Background: Depression imposes a major burden on public health as the leading cause of disability worldwide. Sleep disturbance is a core symptom of depression that affects the vast majority of patients. Nonetheless, it is frequently not resolved by depression treatment and may even be worsened through some pharmaceutical interventions. Disturbed sleep negatively impact patients' quality of life, and persistent sleep disturbance increases the risk of recurrence, relapse, and even suicide. However, the development of novel treatments that might improve sleep problems is hindered by the lack of reliable low-burden objective measures that can adequately assess disturbed sleep in this population.

Summary: Developing improved digital measurement tools that are fit for use in clinical trials for major depressive disorder could promote the inclusion of sleep as a focus for treatment, clinical drug development, and research. This perspective piece explores the path toward the development of novel digital measures, reviews the existing evidence on the meaningfulness of sleep in depression, and summarizes existing methods of sleep assessments, including the use of digital health technologies.

Key messages: Our objective was to make a clear call to action and path forward for the qualification of new digital outcome measures which would enable assessment of sleep disturbance as an aspect of health that truly matters to patients, promoting sleep as an important outcome for clinical development, and ultimately ensure that disturbed sleep will not remain the forgotten symptom of depression.

Introduction: We studied the accuracy of the automatic speech recognition (ASR) software by comparing ASR scores with manual scores from a verbal learning test (VLT) and a semantic verbal fluency (SVF) task in a semiautomated phone assessment in a memory clinic population. Furthermore, we examined the differentiating value of these tests between participants with subjective cognitive decline (SCD) and mild cognitive impairment (MCI). We also investigated whether the automatically calculated speech and linguistic features had an additional value compared to the commonly used total scores in a semiautomated phone assessment.

Methods: We included 94 participants from the memory clinic of the Maastricht University Medical Center+ (SCD N = 56 and MCI N = 38). The test leader guided the participant through a semiautomated phone assessment. The VLT and SVF were audio recorded and processed via a mobile application. The recall count and speech and linguistic features were automatically extracted. The diagnostic groups were classified by training machine learning classifiers to differentiate SCD and MCI participants.

Results: The intraclass correlation for inter-rater reliability between the manual and the ASR total word count was 0.89 (95% CI 0.09-0.97) for the VLT immediate recall, 0.94 (95% CI 0.68-0.98) for the VLT delayed recall, and 0.93 (95% CI 0.56-0.97) for the SVF. The full model including the total word count and speech and linguistic features had an area under the curve of 0.81 and 0.77 for the VLT immediate and delayed recall, respectively, and 0.61 for the SVF.

Conclusion: There was a high agreement between the ASR and manual scores, keeping the broad confidence intervals in mind. The phone-based VLT was able to differentiate between SCD and MCI and can have opportunities for clinical trial screening.

The use of digital phenotyping continues to expand across all fields of health. By collecting quantitative data in real-time using devices such as smartphones or smartwatches, researchers and clinicians can develop a profile of a wide range of conditions. Smartphones contain sensors that collect data, such as GPS or accelerometer data, which can inform secondary metrics such as time spent at home, location entropy, or even sleep duration. These metrics, when used as digital biomarkers, are not only used to investigate the relationship between behavior and health symptoms but can also be used to support personalized and preventative care. Successful phenotyping requires consistent long-term collection of relevant and high-quality data. In this paper, we present the potential of newly available, for approved research, opt-in SensorKit sensors on iOS devices in improving the accuracy of digital phenotyping. We collected opt-in sensor data over 1 week from a single person with depression using the open-source mindLAMP app developed by the Division of Digital Psychiatry at Beth Israel Deaconess Medical Center. Five sensors from SensorKit were included. The names of the sensors, as listed in official documentation, include the following: phone usage, messages usage, visits, device usage, and ambient light. We compared data from these five new sensors from SensorKit to our current digital phenotyping data collection sensors to assess similarity and differences in both raw and processed data. We present sample data from all five of these new sensors. We also present sample data from current digital phenotyping sources and compare these data to SensorKit sensors when applicable. SensorKit offers great potential for health research. Many SensorKit sensors improve upon previously accessible features and produce data that appears clinically relevant. SensorKit sensors will likely play a substantial role in digital phenotyping. However, using these data requires advanced health app infrastructure and the ability to securely store high-frequency data.

Introduction: Technology holds the potential to track disease progression and response to neuroprotective therapies in Parkinson's disease (PD). The sit-to-stand (STS) transition is a frequently occurring event which is important to people with PD. The aim of this study was to demonstrate an automatic approach to quantify STS duration and speed using a real-world free-living dataset and look at clinical correlations of the outcomes, including whether STS parameters change when someone withholds PD medications.

Methods: Eighty-five hours of video data were collected from 24 participants staying in pairs for 5-day periods in a naturalistic setting. Skeleton joints were extracted from the video data; the head trajectory was estimated and used to estimate the STS parameters of duration and speed.

Results: 3.14 STS transitions were seen per hour per person on average. Significant correlations were seen between automatic and manual STS duration (Pearson rho - 0.419, p = 0.042) and between automatic STS speed and manual STS duration (Pearson rho - 0.780, p < 0.001). Significant and strong correlations were seen between the gold-standard clinical rating scale scores and both STS duration and STS speed; these correlations were not seen in the STS transitions when the participants were carrying something in their hand(s). Significant differences were seen at the cohort level between control and PD participants' ON medications' STS duration (U = 6,263, p = 0.018) and speed (U = 9,965, p < 0.001). At an individual level, only two participants with PD became significantly slower to STS when they were OFF medications; withholding medications did not significantly change STS duration at an individual level in any participant.

Conclusion: We demonstrate a novel approach to automatically quantify and ecologically validate two STS parameters which correlate with gold-standard clinical tools measuring disease severity in PD.

Background: Assessment of reliability is one of the key components of the validation process designed to demonstrate that a novel clinical measure assessed by a digital health technology tool is fit-for-purpose in clinical research, care, and decision-making. Reliability assessment contributes to characterization of the signal-to-noise ratio and measurement error and is the first indicator of potential usefulness of the proposed clinical measure.

Summary: Methodologies for reliability analyses are scattered across literature on validation of PROs, wet biomarkers, etc., yet are equally useful for digital clinical measures. We review a general modeling framework and statistical metrics typically used for reliability assessments as part of the clinical validation. We also present methods for the assessment of agreement and measurement error, alongside modified approaches for categorical measures. We illustrate the discussed techniques using physical activity data from a wearable device with an accelerometer sensor collected in clinical trial participants.

Key messages: This paper provides statisticians and data scientists, involved in development and validation of novel digital clinical measures, an overview of the statistical methodologies and analytical tools for reliability assessment.

Introduction: Myasthenia gravis (MG) is a rare autoimmune disease characterized by muscle weakness and fatigue. Ptosis (eyelid drooping) occurs due to fatigue of the muscles for eyelid elevation and is one symptom widely used by patients and healthcare providers to track progression of the disease. Margin reflex distance 1 (MRD1) is an accepted clinical measure of ptosis and is typically assessed using a hand-held ruler. In this work, we develop an AI model that enables automated measurement of MRD1 in self-recorded video clips collected using patient smartphones.

Methods: A 3-month prospective observational study collected a dataset of video clips from patients with MG. Study participants were asked to perform an eyelid fatigability exercise to elicit ptosis while filming "selfie" videos on their smartphones. These images were collected in nonclinical settings, with no in-person training. The dataset was annotated by non-clinicians for (1) eye landmarks to establish ground truth MRD1 and (2) the quality of the video frames. The ground truth MRD1 (in millimeters, mm) was calculated from eye landmark annotations in the video frames using a standard conversion factor, the horizontal visible iris diameter of the human eye. To develop the model, we trained a neural network for eye landmark detection consisting of a ResNet50 backbone plus two dense layers of 78 dimensions on publicly available datasets. Only the ResNet50 backbone was used, discarding the last two layers. The embeddings from the ResNet50 were used as features for a support vector regressor (SVR) using a linear kernel, for regression to MRD1, in mm. The SVR was trained on data collected remotely from MG patients in the prospective study, split into training and development folds. The model's performance for MRD1 estimation was evaluated on a separate test fold from the study dataset.

Results: On the full test fold (N = 664 images), the correlation between the ground truth and predicted MRD1 values was strong (r = 0.732). The mean absolute error was 0.822 mm; the mean of differences was -0.256 mm; and 95% limits of agreement (LOA) were -0.214-1.768 mm. Model performance showed no improvement when test data were gated to exclude "poor" quality images.

Conclusions: On data generated under highly challenging real-world conditions from a variety of different smartphone devices, the model predicts MRD1 with a strong correlation (r = 0.732) between ground truth and predicted MRD1.