Validation of AI Models to Measure Physical Activity After a Stroke

Validation and Testing of Artificial Intelligence Models to Measure Physical Activity in Patients Admitted to Hospital Following a Stroke

The research team are developing algorithms using artificial intelligence that use information collected by accelerometers to detect a person's position, such as whether an individual is lying, sitting, or standing, and the individual's movements, such as whether they are taking steps or standing up.

Sensor location will affect the accuracy of the model and acceptability of the method. The research team are therefore developing algorithms for four different locations.

The purpose of the research is for the development of the algorithms and check whether they accurately recognise different positions and movements in people whose movement is affected by a stroke, and by being in a hospital environment (e.g. using a profiling bed).

The research team plan to recruit between 34 and 50 participants who are admitted hospital due to having a stroke. After providing informed consent, participants will be asked to complete a one-off assessment with a member of the research team and a ward physiotherapist. Participants will be asked to wear the four sensors, and move through a series of postures, walk for up to six minutes, and stand as many times as they feel able in one minute.

Inclusion Criteria:

• admitted to hospital with a diagnosis of an acute stroke.

Exclusion Criteria:

• unable to provide informed consent;
• receiving end-of-life care;
• the consultant in charge of their care disagrees with their inclusion