Spotlight Research: Predicting Therapy Outcomes With Machine Learning
BU Researchers Use Machine Learning to Create Personalized Recovery Plans for Individuals with Neurological Disorders
Boston University Researchers are working to advance personalized healthcare by developing a personalized prediction algorithm for speech, language, and cognitive recovery in patients with neurological disorders. In a project funded by the Hariri Insitute, Boston University Researchers Swathi Kiran, Alan Liu, Claire Cordella, Margrit Betke and Prakash Ishwar, are using a vast mHealth dataset provided by Constant Therapy. This project employs machine learning techniques, with an aim to provide tailored recommendations for patients’ rehabilitation journeys, including information about how often, how much, and for how long an individual should practice to reach desired treatment goals. The findings have the potential to facilitate optimized brain recovery, paving the way for individualized therapy plans that consider diverse patient presentations.
In a significant step towards personalized rehabilitation, this study, conducted by Boston University Researchers analyzes real-world data from over 600,000 users of the Constant Therapy (CT) app. Users of this app often have speech, language, and cognitive difficulties following stroke or other neurological disorders and use the app as a way of rehabilitating functional capabilities across a wide range of domains, such as reading, verbal expression, or auditory comprehension. Because this therapy takes place at-home via an app, patients and clinicians often wish to know how often, how much, and for how long they should practice to meet their goals. The current study aims to answer such questions using cutting-edge machine learning algorithms.
Specifically, researchers are developing machine learning algorithms that utilize user-specific information to predict improvement on functional therapy milestones (e.g., going from reading single words to reading long paragraphs). The development of such algorithms enables the extension to new patients using the app for the very first time. This means that a new user can identify functional skill(s) they desire to improve, provide background demographic information about themselves, and then receive a personalized practice plan suggesting the amount, frequency and duration with which they might practice app-based tasks to reach each functional milestone. The implications of this approach are far-reaching, as the algorithmic approach enables clinicians to provide precise recommendations that cater to each patient’s unique circumstances and goals. This individualized approach allows for a more targeted treatment plan and effective rehabilitation process.
The study represents a promising advancement in the field of neurological rehabilitation, bridging the gap between data-driven insights and personalized therapy. By understanding the nuances of each patient’s condition and progress, healthcare professionals can optimize treatment plans and maximize the chances of successful recovery.
For those interested in exploring the research further, the preliminary study that paved the way for this research can be accessed via the link below.