When looking for treasure, every kid with a treasure map knows that ‘X’ marks the spot. That ‘X’ is a treasure marker and it is very useful information for treasure hunters and pirates alike!
In the medical setting, the maps are much more complicated and often much less complete a vial of blood, a few gigabytes of genomic data, an X-ray, or several questionnaires, just to name a few partial, highly-obscured views of the landscape. For a given malady, researchers explore these types of biological maps, looking for patterns and regularities that are strongly associated with the occurrence of the malady figuring out whether it has a unique signature ‘X’ that marks its presence or progression. When such a data signature is identified and validated, it is called a biomarker.
There are many types of biomarkers at least as many biomarker types as there are biological data types. If the signature is identified in blood, it’s a blood biomarker. If found in genomics data, it’s a genetic biomarker. If it comes from an X-ray or brain scan, it’s an imaging biomarker. And so on! Many of these biomarkers share a few things in common. For example, blood tests, genetic samples, and imaging appointments are often single-point-of-time measurements, or they are measured irregularly, at low frequency (e.g., annually measured, or quarterly at best). The reason for this is that data gathered from these samples is often costly in terms of equipment, in terms of expertise, and in terms of office and lab time.
Not all data is as costly though.
In 2020, most of us have a smartphone in our pocket or bag during significant portions of the day. Some of us have smart watches or fitness trackers. These types of devices mobiles, wearables, generally have an accelerometer onboard, and often a gyroscope and magnetometer as well. These sensor suites cheaply capture data points that inherently carry information regarding your energy levels, motion, and activities, which in turn hold information pertaining to current and evolving biological and psychological states. In other words, it is not impossible to derive important biomarkers from such digital devices.
This is often referred to as a digital biomarker.
At CVB, we explore the sensor data recorded from such mobile and wearable devices, seeking digital biomarkers that can be used to help track the progression (or stasis) of symptoms and conditions that arise in Rett syndrome, Parkinson’s disease, and other brain-related conditions, such as depression and post-traumatic stress disorder.
An important approach in health-tracking analytics is to contextualize the raw sensor data by annotating it with semantically meaningful information, such as whether the subject is currently on the move (active) or idly sitting around (inactive). With some further sophistication, it’s possible to further annotate some basic activities of daily living, such as walking, running, sitting, standing, laying, and so on. Finally, disease-specific behaviors can be detected, classified, and tracked over time as well (e.g., stereotypic hand motions in Rett syndrome, or tremors in Parkinson’s).
The important takeaway from all of this is its profound potential for providing clinicians and other stakeholders with important, time-varying, human-interpretable biomarkers at a relatively high frequency weekly, daily, or even hourly metrics on a screen, showing clear trends and fluctuations. For some diseases, this means that a clinician no longer has to rely on scattershot journal notes and patient questionnaires every few months to infer whether a new medical treatment plan is working or not. For others, this means that expensive, low frequency biomarkers can be supported by and at times, possibly even replaced by cheap, high frequency proxies.
Probably the most important implication of digital biomarkers derived from commonly used devices, such as mobile phones, is that the discovery and publishing of such biomarkers is open to the world, allowing individuals to better track and understand their own health and healthcare.