Back pain management: a personalised smartphone-based solution
Engineering, medical and health science researchers have been collaborating with the medical device industry to study low back pain via smartphones and wearables. The goal was to improve understanding of low back pain and to develop new solutions for therapy and prevention using mobile health technologies.
Portrait / project description (completed research project)
An engaging mHealth app will enable people to assess and document their activities related to low back pain (LBP). This will allow to collect important data from a group of people at risk of suffering from back pain and will shed light on people’s individual experience of low back pain. New experimental strategies should be developed in this project to test different methods of preventing or reducing low back pain, such as physiotherapy, education, and mobile games for health. With small groups of app users these methods should be evaluated for effectiveness. In addition to determining which treatments work and why, it should also be investigated how to select the most suitable method for each user based on his or her personal data profile.
Low back pain affects almost all individuals in Switzerland at least once at some point in their lives. It is a complex illness whose cause is not fully understood. Costly surgery is undesirable, but the variable nature of the illness makes it challenging for doctors to find effective treatments. A key step in solving the problem is to better understand the needs of individuals during episodes of low back pain and their responses to various treatments. Digital technologies such as smartphones, wearable sensors and machine learning make this approach possible.
The objective of the project was to use novel mobile health (mHealth) tools for identifying changes in low back pain among people in Switzerland. Machine learning models should be developed for personalising ways to prevent low back pain using a large set of data recorded with smartphone apps from people with and without low back pain at home and in the clinic. Novel strategies for making health data meaningful, developing analysis software, and handling this sensitive health data should be introduced.
This project will provide clinicians with new mHealth technologies to track progress and enhance understanding of low back pain. Data scientists working with health data will benefit from new software and algorithms that simplify data collection, ensure high data quality and provide accurate analysis. Finally, society will benefit from useful health insights and novel solutions for better management of back pain.
The smartphone app “Swiss Health Challenge” which allowed to collect, transmit and store anonymised data for research was developed in this project. The app was designed to engage with the user in form of challenges. It was used in two clinical studies where former patients and persons at risk for LBP were investigated.
The first trial provided new insights about the use of digital tools for treatment support. The use of the exergaming intervention did not affect the movement outcomes. This result confirms that a population who is not severely affected by LBP is difficult to monitor and treat. However, the study stands out from others in the use of digital tools for objectively monitoring of adherence in real-life settings. Also, the results provided new insights that fear of movement was related to several direction specific parameters of postural sway in LBP.
In the second study adolescent skiers were monitored to identify overstress that is potentially leading to back injury. The app was used to monitor ski training with the altitude and acceleration sensors. It could be shown that this information can provide far more details than self-reports.
To process all this big medical data, a series of machine learning tools that advanced the field was developed:
- Introduction of a new model to integrate arbitrary numbers of symptom assessments from multiple types of measurements over long periods of time.
- Development of a mechanism that quantifies the importance of the measurements.
- Suggestion of new methods to perform the learning tasks much faster than existing end-to-end deep learning approaches.
- Demonstration that the project’s approaches lead to significant improvements in prediction performance. With these methods the challenges of missingness, sparse data, long-term temporal dependencies between measurements, and multivariate data with irregular sampling could be solved.
Despite new insights from the studies and advancements, the results acknowledge that LBP remains a complex and poorly understood disease that requires further research and investigation.
Personalized management of low back pain with mHealth: Big Data opportunities, challenges and solutions