- March 19, 2013
- 2:19 pm
- Konstantin Läufer
Loyola CS Colloquium Fri 22 Mar 2013 1:15-2:15 pm LT-412 WTC
Measuring Patient Mobility with Mobile Phones through Machine Learning
Mark V. Albert, Ph.D.
Center for Bionic Medicine
Rehabilitation Institute of Chicago
Fri 22 Mar 2013 1:15-2:15 pm
Lewis Towers LT-412 (Linux Lab)
Water Tower Campus
Loyola University Chicago
820 N Michigan Ave (use 111 E Pearson St entrance)
For patients with disorders that affect movement, including stroke, heart disease, spinal cord injury, and Parkinson’s disease, establishing the efficacy of different drug dosages and physical therapy regiments can be difficult. Most measures require time-consuming clinical visits, costly additional equipment, or tedious journaling of symptoms. Fortunately, many people already carry all the equipment necessary to conveniently and objectively infer their health and mobility throughout the day, with no additional cost.
In my research projects at the Rehabilitation Institute of Chicago (RIC) I have been using sensors in mobile phones to track patient progress, recognize their activities, and infer their functional capabilities for prosthetics. Using machine learning, I have analyzed accelerometer signals from patients with Parkinson’s disease to identify their specific activities; this is the first step in an ongoing effort to passively measure symptoms throughout the day. I have applied a similar machine learning strategy to greatly improve fall detection and classification; falls are a common occurrence in patients with motor disorders, and fall prevention studies can benefit from this simpler, more objective data collection. In order to better match prosthetic legs to patients, I have shown how passively measured activity levels can be related to medicare-based functional classes of patients. Also, I will discuss the role of my remote sensing system in the RIC outcomes dashboard – a pilot clinical program that won last year’s Henry Betz innovation award; the project produced an interface that allowed clinicians to view the change in movements and clinical measures over time for an entire floor of patients. For these patients, this work influenced the decision making process that is part of the standard of care at RIC.
Ultimately, the goal of my work is to promote the use of low-cost sensors and mobile applications to track disease progression and the effect of therapies through applied machine learning and improved interface design. Such tracking enables the enrollment of larger numbers of patients over extended periods of time with greater convenience for participants. Such convenient, objective, and continuously recorded information will be invaluable to researchers and clinicians in improving the care of patients with impairments in mobility.
Dr. Albert is a postdoctoral research associate at the Rehabilitation Institute of Chicago and Northwestern University. His current research applies machine learning to automatically interpret data collected from remote sensors carried by patients, including the accelerometers in their mobile phones. He has published a number of papers measuring motor adaptation, detecting falls, and recognizing patient activities.
Prior to Chicago, he received his Ph.D. in computational biology from Cornell University, with an emphasis on applying efficient coding principles to computational neuroscience. While a graduate student there, he was also part of the mobile phone-based startup company Instinctiv. Before Cornell, as research assistant at Carnegie Mellon University, he applied computational models to cognition through human fMRI experiments and to vision in primate neurophysiology. He is a Fulbright Scholar from the University of Vienna and a graduate from Pittsburgh State University.
via Tumblr http://blog.cs.luc.edu/post/45770504847