Digital endpoints for self-administered home-based functional assessment in pediatric Friedreich's ataxia.

BACKGROUND: Friedreich's ataxia is an inherited, progressive, neurodegenerative disease that typically begins in childhood. Disease severity is commonly assessed with rating scales, such as the modified Friedreich's Ataxia Rating Scale, which are usually administered in the clinic by a neurology specialist. OBJECTIVE: This study evaluated the utility of home-based, self-administered digital endpoints in children with Friedreich's ataxia and unaffected controls and their relationship to standard clinical rating scales. METHODS: In a cross-sectional study with 25 participants (13 with Friedreich's ataxia and 12 unaffected controls, aged 6-15 years), home-based digital endpoints that reflect activities of daily living were recorded over 1 week. Domains analyzed were hand motor function with a digitized drawing, automated analysis of speech with a recorded oral diadochokinesis test, and gait and balance with wearable sensors. RESULTS: Hand-drawing and speech tests were easy to conduct and generated high-quality data. The sensor-based gait and balance tests suffered from technical limitations in this study setup. Several parameters discriminated between groups or correlated strongly with modified Friedreich's Ataxia Rating Scale total score and activities of daily living total score in the Friedreich's ataxia group. Hand-drawing parameters also strongly correlated with standard 9-hole peg test scores. INTERPRETATION: Deploying digital endpoints in home settings is feasible in this population, results in meaningful and robust data collection, and may allow for frequent sampling over longer periods of time to track disease progression. Care must be taken when training participants, and investigators should consider the complexity of the tasks and equipment used.

HistoNet: A Deep Learning-Based Model of Normal Histology.

We introduce HistoNet, a deep neural network trained on normal tissue. On 1690 slides with rat tissue samples from 6 preclinical toxicology studies, tissue regions were outlined and annotated by pathologists into 46 different tissue classes. From these annotated regions, we sampled small 224 × 224 pixels images (patches) at 6 different levels of magnification. Using 4 studies as training set and 2 studies as test set, we trained VGG-16, ResNet-50, and Inception-v3 networks separately at each magnification level. Among these model architectures, Inception-v3 and ResNet-50 outperformed VGG-16. Inception-v3 identified the tissue from query images, with an accuracy up to 83.4%. Most misclassifications occurred between histologically similar tissues. Investigation of the features learned by the model (embedding layer) using Uniform Manifold Approximation and Projection revealed not only coherent clusters associated with the individual tissues but also subclusters corresponding to histologically meaningful structures that had not been annotated or trained for. This suggests that the histological representation learned by HistoNet could be useful as the basis of other machine learning algorithms and data mining. Finally, we found that models trained on rat tissues can be used on non-human primate and minipig tissues with minimal retraining.

Biomarker-Based Classification and Localization of Renal Lesions Using Learned Representations of Histology-A Machine Learning Approach to Histopathology.

Several deep learning approaches have been proposed to address the challenges in computational pathology by learning structural details in an unbiased way. Transfer learning allows starting from a learned representation of a pretrained model to be directly used or fine-tuned for a new domain. However, in histopathology, the problem domain is tissue-specific and putting together a labelled data set is challenging. On the other hand, whole slide-level annotations, such as biomarker levels, are much easier to obtain. We compare two pretrained models, one histology-specific and one from ImageNet on various computational pathology tasks. We show that a domain-specific model (HistoNet) contains richer information for biomarker classification, localization of biomarker-relevant morphology within a slide, and the prediction of expert-graded features. We use a weakly supervised approach to discriminate slides based on biomarker level and simultaneously predict which regions contribute to that prediction. We employ multitask learning to show that learned representations correlate with morphological features graded by expert pathologists. All of these results are demonstrated in the context of renal toxicity in a mechanistic study of compound toxicity in rat models. Our results emphasize the importance of histology-specific models and their knowledge representations for solving a wide range of computational pathology tasks.

Robust Step Detection from Different Waist-Worn Sensor Positions: Implications for Clinical Studies.

Analyzing human gait with inertial sensors provides valuable insights into a wide range of health impairments, including many musculoskeletal and neurological diseases. A representative and reliable assessment of gait requires continuous monitoring over long periods and ideally takes place in the subjects' habitual environment (real-world). An inconsistent sensor wearing position can affect gait characterization and influence clinical study results, thus clinical study protocols are typically highly proscriptive, instructing all participants to wear the sensor in a uniform manner. This restrictive approach improves data quality but reduces overall adherence. In this work, we analyze the impact of altering the sensor wearing position around the waist on sensor signal and step detection. We demonstrate that an asymmetrically worn sensor leads to additional odd-harmonic frequency components in the frequency spectrum. We propose a robust solution for step detection based on autocorrelation to overcome sensor position variation (sensitivity = 0.99, precision = 0.99). The proposed solution reduces the impact of inconsistent sensor positioning on gait characterization in clinical studies, thus providing more flexibility to protocol implementation and more freedom to participants to wear the sensor in the position most comfortable to them. This work is a first step towards truly position-agnostic gait assessment in clinical settings.

Continuous Digital Monitoring of Walking Speed in Frail Elderly Patients: Noninterventional Validation Study and Longitudinal Clinical Trial.

BACKGROUND: Digital technologies and advanced analytics have drastically improved our ability to capture and interpret health-relevant data from patients. However, only limited data and results have been published that demonstrate accuracy in target indications, real-world feasibility, or the validity and value of these novel approaches. OBJECTIVE: This study aimed to establish accuracy, feasibility, and validity of continuous digital monitoring of walking speed in frail, elderly patients with sarcopenia and to create an open source repository of raw, derived, and reference data as a resource for the community. METHODS: Data described here were collected as a part of 2 clinical studies: an independent, noninterventional validation study and a phase 2b interventional clinical trial in older adults with sarcopenia. In both studies, participants were monitored by using a waist-worn inertial sensor. The cross-sectional, independent validation study collected data at a single site from 26 naturally slow-walking elderly subjects during a parcours course through the clinic, designed to simulate a real-world environment. In the phase 2b interventional clinical trial, 217 patients with sarcopenia were recruited across 32 sites globally, where patients were monitored over 25 weeks, both during and between visits. RESULTS: We have demonstrated that our approach can capture in-clinic gait speed in frail slow-walking adults with a residual standard error of 0.08 m per second in the independent validation study and 0.08, 0.09, and 0.07 m per second for the 4 m walk test (4mWT), 6-min walk test (6MWT), and 400 m walk test (400mWT) standard gait speed assessments, respectively, in the interventional clinical trial. We demonstrated the feasibility of our approach by capturing 9668 patient-days of real-world data from 192 patients and 32 sites, as part of the interventional clinical trial. We derived inferred contextual information describing the length of a given walking bout and uncovered positive associations between the short 4mWT gait speed assessment and gait speed in bouts between 5 and 20 steps (correlation of 0.23) and longer 6MWT and 400mWT assessments with bouts of 80 to 640 steps (correlations of 0.48 and 0.59, respectively). CONCLUSIONS: This study showed, for the first time, accurate capture of real-world gait speed in slow-walking older adults with sarcopenia. We demonstrated the feasibility of long-term digital monitoring of mobility in geriatric populations, establishing that sufficient data can be collected to allow robust monitoring of gait behaviors outside the clinic, even in the absence of feedback or incentives. Using inferred context, we demonstrated the ecological validity of in-clinic gait assessments, describing positive associations between in-clinic performance and real-world walking behavior. We make all data available as an open source resource for the community, providing a basis for further study of the relationship between standardized physical performance assessment and real-world behavior and independence.

Validity of accelerometry in step detection and gait speed measurement in orthogeriatric patients.

BACKGROUND: Mobile accelerometry is a powerful and promising option to capture long-term changes in gait in both clinical and real-world scenarios. Increasingly, gait parameters have demonstrated their value as clinical outcome parameters, but validation of these parameters in elderly patients is still limited. OBJECTIVE: The aim of this study was to implement a validation framework appropriate for elderly patients and representative of real-world settings, and to use this framework to test and improve algorithms for mobile accelerometry data in an orthogeriatric population. METHODS: Twenty elderly subjects wearing a 3D-accelerometer completed a parcours imitating a real-world scenario. High-definition video and mobile reference speed capture served to validate different algorithms. RESULTS: Particularly at slow gait speeds, relevant improvements in accuracy have been achieved. Compared to the reference the deviation was less than 1% in step detection and less than 0.05 m/s in gait speed measurements, even for slow walking subjects (< 0.8 m/s). CONCLUSION: With the described setup, algorithms for step and gait speed detection have successfully been validated in an elderly population and demonstrated to have improved performance versus previously published algorithms. These results are promising that long-term and/or real-world measurements are possible with an acceptable accuracy even in elderly frail patients with slow gait speeds.

Continuous Monitoring of Patient Mobility for 18 Months Using Inertial Sensors following Traumatic Knee Injury: A Case Study.

Continuous patient activity monitoring during rehabilitation, enabled by digital technologies, will allow the objective capture of real-world mobility and aligning treatment to each individual's recovery trajectory in real time. To explore the feasibility and added value of such approaches, we present a case study of a 36-year-old male participant monitored continuously for activity levels and gait parameters using a waist-worn inertial sensor following a tibial plateau fracture on the right side, sustained as a result of a high-energy trauma during a sporting accident. During rehabilitation, data were collected for a period of 553 days, with > 80% daytime compliance, until the participant returned to near full mobility. The participant completed a daily diary with the annotation of major events (falls, near falls, cycling periods, or physiotherapy sessions) and key dates in the patient's recovery, including medical interventions, transitioning off crutches, and returning to work. We demonstrate the feasibility of collecting, storing, and mining of continuous digital mobility data and show that such data can detect changes in mobility and provide insights into long-term rehabilitation. We make both raw data and annotations available as a resource with the aspiration that further methods and insights will be built on this initial exploration of added value and continue to demonstrate that continuous monitoring can be deployed to aid rehabilitation.