Behavioral Pattern Recognition through Wearable Devices

Goal

Develop digital biomarkers to better understand human behavioral patterns related to indications such as: depression, chronic pain, etc.

Method

Mobile phone health apps are increasingly gaining attention in various therapeutic areas. There is a growing interest in using smartphone sensing to infer human dynamics and behavioral health. The widespread and everyday use of smartphones can be harnessed as an instrument for non-intrusive monitoring of several behavioral indicators in the field of mental health, chronic pain-related diseases etc.

Smartphone data, such as: activity data, app usage, call duration, screen on and of times, voice data etc., can be leveraged for close monitoring and assessment at scale, that exceeds what is feasible with the currently existing in-clinic assessments.

	Clinical Trials - The data is collected by running clinical trials in collaboration with leading medical centers/hospitals. The study cohorts are derived from the indications the study aims to explore (depression, chronic pain etc.).
	Patient recruitment - Patients are recruited by the physicians in the medical centers/clinics/hospitals. The number of patients and the cohorts are determined according to the study protocol. Each patient installs two apps on their personal phone – a passive data collection app and a voice recognition app.
	Passive data collecting app (iFeel) - Data collected from phone's sensors include screen and device on/off times, calls and texts logs, apps usage information, location, luminosity, acceleration and more. By using a passive data collection app, the burden on the patients is minimal, and we avoid possible biases during the clinical study.
	Voice recognition app (iHear) – Some of the patients install a voice recognition app on their phones. The voice recording records answers to predefined questions, where the goal is to capture not just the content of the answer, but also the voice trait of how the answer was delivered, both in terms of the selected words and phrases, but also the vocal characteristics (e.g. prosody).
	Data analysis – Descriptive analytics, machine learning and statistical analysis is employed in order to develop and validate digital markers such as sleep disturbance, physical activity and mobility, social anxiety, daily life routine and more. Our aim is to provide an unprecedented wealth of information about digital parameters that are relevant to behavioral patterns. Scientific analysis of this data would lead to new insights and may pave the way to form objective measurements that can be used to enhance treatment as well as to predict outcomes and progression, and to assess risks.

Behavioral Pattern Recognition through Wearable Devices

Goal

Method

Resources