Automated Quality Control for Sensor Based Symptom Measurement Performed Outside the Lab

Abstract

The use of wearable sensing technology for objective, non-invasive and remote clinimetric testing of symptoms has considerable potential. However, the accuracy achievable with such technology is highly reliant on separating the useful from irrelevant sensor data. Monitoring patient symptoms using digital sensors outside of controlled, clinical lab settings creates a variety of practical challenges, such as recording unexpected user behaviors. These behaviors often violate the assumptions of clinimetric testing protocols, where these protocols are designed to probe for specific symptoms. Such violations are frequent outside the lab and affect the accuracy of the subsequent data analysis and scientific conclusions. To address these problems, we report on a unified algorithmic framework for automated sensor data quality control, which can identify those parts of the sensor data that are sufficiently reliable for further analysis. Combining both parametric and nonparametric signal processing and machine learning techniques, we demonstrate that across 100 subjects and 300 clinimetric tests from three different types of behavioral clinimetric protocols, the system shows an average segmentation accuracy of around 90%. By extracting reliable sensor data, it is possible to strip the data of confounding factors in the environment that may threaten reproducibility and replicability.

DAQUA-MASS: An ISO 8000-61 Based Data Quality Management Methodology for Sensor Data

“…In this paper, authors address the problem of uncertainty of data coming from sensors with an approach dedicated to providing environmental monitoring applications and users with data quality information. Badawy et al [ 26 ] combined parametric and non-parametric signal processing and machine learning algorithms for automating sensor data quality control, which can identify those parts of the sensor data that are sufficiently reliable for further analysis and discards useless data.…”

Quality Control of Voice Recordings in Remote Parkinson’s Disease Monitoring Using the Infinite Hidden Markov Model

“…We repeated the CV procedure 200 times to obtain the distribution of classification accuracies. We compare the proposed iHMM approach trained in the MFCC domain with two different baseline methods: the energy-based VAD which computes the energies of all frames, selects the maximum, and then sets the detection threshold as 30 dB below the maximum [27]; and the nonparametric switching AR model proposed in [14] which takes as an input the energy of each frame. We refer to the former baseline as the VAD-based method and to the latter one as NPSARbased method in the rest of the paper.…”

“…Additive smoothing techniques [26], in which a small sample-correction is added to all probability estimates, are popular approaches to prevent the probability estimate to be zero. In [14], an unobserved state indicator is classified as protocol violation. In this paper, we take advantage of the MFCC's properties to tackle this problem.…”

“…For example, talking, laughing or coughing may all be considered as protocol violations in sustained vowel data sets even though the recordings are of high quality. Badawy et al [14] proposed a general framework for detecting a wide range of protocol violations using a nonparametric switching autoregressive (AR) model; however, low order AR models fail to capture variations in the low frequency harmonics.…”

Probabilistic Modelling of Gait for Robust Passive Monitoring in Daily Life

“…Low pass filters can be used in an accelerometer only setup, but they are poorly justified, since orientation changes can have a broad bandwidth leading to unwanted distortions in the time domain depending on the cut off frequency of the filter. In this work we opt for a piecewise l 1 -trend filter as motivated in Badawy et al [44] which assumes that changes due to orientation are piecewise linear [45].…”