MP: A Smart Mobile App to Facilitate Rehabilitation of Stroke Patients

Stroke is a common cause of disability, and many persons with stroke receive physiotherapy over many years. This project aims to develop a mobile app (SmartRehab) that makes cooperation between persons with stroke and their physiotherapists easier. The app, combined with small movement sensors, will make it possible to self-monitor daily activities and provide feedback about performance during home exercise. Wearable sensors have become increasingly popular to monitor physical activity and motivate people towards a more active lifestyle. However, commercial wearables are not suited to detect the asymmetrical movement patterns of people with stroke. In our project, we use machine learning algorithms to address the limitations of existing apps in the market. The product we are developing is a mobile app combined with wearable sensors that allow the collection of movement data of persons with stroke and to classify and quantify activities related to gait and mobility. SmartRehab helps persons with stroke self-monitor their daily activities and supports the physiotherapists to provide better guidance during home-based training, thereby follow-up therapist-supervised training. Before this project, SmartRehab was tested with healthy students. In this project, we developed this product further in collaboration with therapists. We evaluated the activity classification accuracy of the improved version of the SmartRehab app with 16 stroke patients. The evaluation happened in two days. On the first day, the patients wore the sensors and performed typical rehabilitation activities. The sensors recorded the movement data and used the data to train the machine learning model in SmartRehab. On the second day, the patients wore the sensors but performed rehabilitation activities like in a free-living environment. SmartRehab uses the trained machine learning model to classify the activities and count the quantity of each activity. The therapists guided the patients on both days. We also performed focus group interviews with 12 stroke patients and 11 physiotherapists to collect their opinions about expected benefits or potential barriers to using this kind of wearable technology. Results show that the machine learning algorithm could classify all the activities precisely on day one, using 80% of sensor data as the training dataset and 20% of the data as the testing dataset. The algorithm could also classify some activities, such as walk-sideways, fairly precisely using day two data. However, the model needs to be improved to recognise some other activities more precisely. The focus group interview results show that a well-designed wearable system has the potential to increase stroke survivors' motivation for home exercise. Wearable technology that intends to give feedback on home exercises should be highly flexible and adaptable to the individual needs of the users. An app that can implement rehabilitation goals and share exercise data with the physiotherapists' journal system can be useful in rehabilitation. The app cannot replace the cooperation between the stroke survivor and their physiotherapist but has the potential to promote it.

The project results include three elements. First, we have developed the SmartRehab app in the future by collaborating with therapists. The new version of the app is more suitable to be used by stroke survivors in their home exercises. Second, we have better understood the advantages and limitations of the machine learning algorithm based on evaluating it with stroke patients. The insights can help us improve the algorithm and the SmartRehab app in the follow-up projects. Third, the focus group interviews with therapists and stroke patients confirmed the app's expected benefits and gave information on new features to be added to the app. Results of the project bring actual and potential technological, societal, economic, and environmental impacts. Through scientific publication and knowledge dissemination to the healthcare research community, the technical impact is that the community increases knowledge about the benefits and requirements of the wearable system for motivating and facilitating stroke survivors' home rehabilitation. There are millions of first-time strokes each year worldwide and about ten thousand in Norway. Recovery from stroke may take weeks, months, or even years. Many patients with minor and moderate disabilities need rehabilitation at home. The SmartRehab app can save the patients' and society's medical care expenses by facilitating home rehabilitation. The app can also reduce the patients' travels to the rehabilitation centres and their therapists.