Multicenter Osteoarthritis Study
The Multicenter Osteoarthritis Study, or MOST, is a unique, large, long-term observational study of risk factors for knee OA. MOST includes patient metadata, clinical outcome data, accelerometer data, and ground reaction force data for approximately 4000 participants. Due to the wealth of data provided by MOST, there are many paths of analysis. Below are two directions that we believe will be the most fruitful.
Physical Activity as a Risk Factor
Mechanical loading is a well-studied risk factor for knee osteoarthritis (OA). However, much of the research has been performed in a controlled laboratory setting and does not capture the day-to-day variety of forces knees experience. Our objective is to predict knee OA progression evaluated by various clinical outcomes (cartilage damage, pain, etc.), from 7 days of accelerometer data using artificial intelligence and deep learning to develop and explain relationships between physical activity and knee OA.
Raw Ground Reaction Force Waveforms to Predict Knee OA
Ground reaction forces represent the total resultant force felt at the foot contact point with the ground. However, to measure mechanical loading at the knee joint we need kinematics captured in a laboratory setting using motion capture. Despite this, knee loading is embedded in the original ground reaction force signal. Our objective is to predict knee OA progression evaluated by various clinical outcomes (cartilage damage, pain, etc.), from raw ipsilateral ground reaction forces without kinematic data using fused multimodal deep learning models.
Florida Moves (FLoMo) Study
Currently, we are collecting data on how people walk and perform daily activities (i.e., sit in a chair, step-up, etc.) in order to create a large data bank of human kinematics and kinetics. With this data, which includes healthy participants, and participants with knee osteoarthritis, we can use machine learning techniques to determine if there are movement patterns associated with faster progression of this degenerative disease.
Pain Flares
Coming Soon!
Shared Strides Study
New markerless motion capture utilizes video data and deep learning to provide biomechanical data in much less time than traditional methods. In addition, data collections in community locations reduce barriers to study participation and may increase diversity of study samples when compared to lab-based settings. The goals of this pilot project are to compare feasibility, acceptability, and participant retention-rates of community-based vs. traditional lab-based data collections.
Funding: Pepper Center