CHIMERA seminar with Professor Derek Hill

Please note, this is a hybrid event. You are welcome to join in-person (UCL staff and students only) or online.

It is often proposed that tracking technology  can help better understand the symptoms of people living with chronic illness, and how well they can live independently, and pick up early warning signs of risks that might need proactive management.  There are two common ways of tracking activities of patients over time in their homes:

• to detect where they are for example, with fixed motion detectors (eg: PIR or ToF).
• To assess their movement using wearable motion sensors (eg: accelerometer, sometimes with addition of gyroscope, magnetometer or pressure sensor)

Fixed motion sensors can measure, for example, visits to the bathroom or kitchen, and how many times people move between rooms, but are limited in their ability to distinguish between people in multi-person households, or whether someone is being active or sedentary in a given room.  Motion sensors can measure activity over time, but can’t localise this without the house.

We have implemented a method that combines both: a wearable bracelet sensor that contains its own sensors and can also receive information from room-mounted sensors, using technology developed by Panoramic Digital Health. The data collected includes the movement sensor data from accelerometer, magnetometer and barometric pressure sensor in the bracelet, labelled with the signal strength of in-room Bluetooth beacons (and values of environment sensors they contain).  The combination of these data provides a signature that combines both position and physical activity information, which we believe is potentially better able to capture meaningful measures of function than either location or activity alone.  We are running three studies collecting this data in a research setting:

•  We also have a validation study in which individual used the technology in a “living lab” with video surveillance providing a “gold standard” reference
• patients with chronic fatigue symptoms (and healthy controls) who also provide a numeric fatigue score several times per day
• pairs of co-habiting individuals, where we try to look for differences in activities and behaviours depending on whether the individuals are together or apart.

In this presentation, I will show examples of the data, and the potential value of these combined measures.

We are looking for data scientists who are interested in considering how to analyze this data in more sophisticated ways to the basic ones we are currently using. We can provide anonysed access to this data for anyone who would like to do some timeseries analysis of the combined activity and location data and look for signatures that  may help us interpret this rich source of data.