SPECIAL SYMPOSIUM: Using Machine Learning and Wearable Devices to Advance Research and Patient Care Using Spinal Cord Injury as a Model SS12

A core tenet of biomedical research is to gain a better understanding of diagnosing, preventing, and treating diseases in humans. The pursuit of this knowledge has been hampered by the available technology, techniques, and analytical approaches to complex data. In recent years the growth of Artificial Intelligence, and the subcategory of machine learning(ML) has increased our ability to analyze data.

The technological advances in research grade and consumer grade wearable devices have improved our ability to make measurements in real-time in everyday life. The combination of ML and wearable devices offers a tremendous capacity to collect and analyze new data, as well as refute or support anecdotes used to influence care. Moreover, the use of wearable technology and ML may facilitate what expert clinicians see in therapy that may not be captured in rigorous research study, augmenting our understanding and interpretation of clinical trials. This information can be used to alter health related decision making and facilitate personalized medicine.

We will highlight the importance of understanding the wearable devices, from data sets, participant burden, device limitations, and reliability and validity. We will provide a foundational understanding of statistical commonality, algorithm types, types of bias and confounding, and real-world considerations. We will discuss how these factors, along with designing a researcher question, ultimately lead to a decision matrix that influences the design of a research study for wearable devices and ML. Finally, we will discuss obstacles involving buzzwords that may impact the dissemination of these data.

All of these factors will be discussed in the context of designing a study for those with spinal cord injury(SCI). We elected to use SCI as a model due to the unique autonomic consequences that are known to follow SCI, and have direction implications for participant recruitment, devices selection, outcome selection, and interpretation of results.