Contact: Dr. Neal Lathia, Computer Lab
Ongoing health conditions are generally managed via intermittent visits to doctors, which works well for detecting gradual changes, but will not pick up rapid change. Similarly, these visits cannot detect changes to a person’s behaviour which may make them at higher risk of either relapsing and developing a new condition.
The researchers therefore decided to look at how smartphones could be used to support and improve this process, using a combination of passive data from the phone (GPS, accelerometer, app usage, call/text patterns etc), and user-completed surveys. They did this by developing a framework for collecting the data, and then used machine-learning techniques to interpret the information as a predictor of different symptoms.
An initial version of the software has been released as a consumer app called EmotionSense which has had over 35,000 downloads, and typically has 2,000 active monthly users at any time. EmotionSense collects the passive data from an Android smartphone, and uses it to measure the user’s mood and happiness.
The researchers are also collaborating with more medically-focused trials of the technology, including a study with Addenbrookes Hospital where an app called QSense is assisting in smoking cessation programmes by detecting trigger behaviours and locations.
The market for healthcare apps is moving fast, and there are many companies working in the consumer health area. The challenge for the i-Team is to look at where a machine-learning based approach can add the most value, and to determine which areas would be the best for the researchers to focus on going forward.