Using Consumer Biometric Sensors for Physiatric Research: An Introduction

Research in using technology for rehabilitation is a burgeoning field. Many exciting high-tech devices such as bionic limbs and exoskeletons are in development with funding from large grant sources such as the Defense Advanced Research Projects Agency. However, there has also been an explosion in development of consumer-grade biometric sensors that have the potential for great utility in the rehabilitation environment as well.

These relatively low-cost commercial products offer free, easy-to-use programming libraries to access sensor data that can be used to create custom research projects. Some products are cheap enough that one can create proof of concepts and even pilot studies with little-to-no dependence on grant awards. Besides smartphones with standard accelerometers and pressure sensors, other new devices offer electromyography, electroencephalography, eye tracking, virtual reality, and more.

Microsoft Kinect Sensor for Xbox One (Wikipedia)

There has been a paradigm shift in the industry — rather than manufacture closed, proprietary hardware and software designed for specialized uses, companies now produce hardware with open access to data that can be used in a variety of fields. An early example of the shift is the Microsoft Kinect, an accessory that detects user’s joint and limb positions in space. It was originally an exclusive item for Xbox gaming consoles but is now openly supported by Microsoft for use by anyone with a Windows PC. Many newer consumer products have launched with similar open data support.

The way devices offer access to their data is via an application program interface (API). The commands generally provide high-level information that has been filtered and processed already. For example, FitBit and iPhone will provide “number of steps” rather than the raw accelerometer force readings that define a “step.” The hard part of the analysis is conveniently done, but the algorithms are usually proprietary. Thus, it is not always clear exactly how the high level data are generated, although some devices do provide raw data output as an option.

Generally, biometric sensors function as follows: sensors within the device (such as accelerometers, capacitive touch, bare electrodes, or infrared detectors) are sampled thousands of times per second to produce raw data. This data can be transferred to a computer for processing, or the data can undergo software analysis on the device and be stored as representations with higher-level meaning such as steps, pulse, or gestures. Data is usually transferred to a phone or laptop wirelessly via Bluetooth or wired via Universal Serial Bus (USB).

The Myo armband provides surface electromyography via Bluetooth. Source: Thalmic Labs screenshot

There are plenty of questions to consider regarding the fleet of new low-cost devices. How accurate are the data they produce? What clinical information does one hope to gather from a device, and can that information be obtained even if a device may not provide the level of accuracy and precision of its laboratory-grade alternatives? What is the best way to filter noise from the flood of biometric data points, and how can the data be further integrated to provide meaningful information to a clinician? Further research is needed, and it is feasible that short-term pilot studies can be conducted within the schedules of resident and fellow physiatrists.

To promote this initiative, I would like to develop a series of tutorials to create simple projects using different biometric sensors with a physiatric focus. My goals are: (1) to review what devices exist; (2) to suggest physiatric use cases and ideas for these devices; (3) to provide walk-through tutorials for building software to interface with data; and (4) to establish a community online of physiatrist builders for the exchange of ideas, support, and research collaboration.

In addition to technical tutorials, I would like to curate “case studies” in using technology with patients. These articles would focus on the experiences of using available technology in the rehab setting rather than technical development of software/hardware from scratch. For example, what commercial equipment is used? What are the upfront and operational costs involved? What is the training regimen like for patients and staff? What is the author’s role in the setup? What are the advantages and disadvantages identified after using the system?

All of these articles would appear online as part of physiatry.org and some may be featured in future Physiatry in Motion newsletters. If anyone is interested in contributing to tutorials or case studies, please contact me!


George Marzloff, MD is a resident in the Department of Rehabilitation Medicine at the Icahn School of Medicine at Mount Sinai.

This article originally was published on April 27, 2016 in Physiatry in Motion, the newsletter published by the Association of Academic Physiatrists Residents’ and Fellows’ Council.