Here is a quick description of the projects I am most excited about that I am working on today in artificial intelligence and big data, and projects that I think will have the biggest impact that I have worked on in the past (with links to the published articles). Please reach out with any questions or interest in collaboration!
Current Highlighted Work
What are the genetic factors associated with back pain, surgery for back pain, and adult scoliosis?
Can you predict the onset of back pain, the need for surgery for back pain, and the onset of scoliosis using genes? What genes increase your risk for spinal stenosis?
Can we use comptuer vision to analyze if a spinal fusion has healed after surgery?
After spinal fusion surgery, a CT scan is often needed to evaluate if a spine has fused. Can we use computer vision to do find an answer based on XR?
How many images are necessary to train a classification algorithm on medical images?
How many images are needed to build an algorithm that can classify an XR. For example, is 100, 200, 500, or a 1000 images needed to train an algorithm to diagnose OA from an XR?
Highlighted Recent Work (Last 3 Years)
Right now, a ton of data about our patients is hiding in our clinical and operative notes. We used novel natural language processing techniques to collect important info about patient historical and exam information, what was done in the operating room, and its effect on their clinical outcomes. This will allow for creation of large surgical databases across hospitals which should improve the quality of care.
The outcome that matters most to patients is not how they look on X-Ray, but how them report their health in a patient reported to outcome. We found that AI algorithms can predict changes in PROs after an injury or sugery which can help treatment planning
After joint replacement surgery, your joint is at risk of loosening and needing replacement. Currently, there are no good methods to know if your joint is loose. We created a novel image recognition algorithm to predict a loose joint from a simple X-Ray.
It has been shown that having a doctor who is the same race or gender as yourself is important in primary care and can improve outcomes. Is this concordance also important in orthopedic care, where gender and racial diversity is lacking?
We have shown in the past how AI/ML can be used to segment an MRI image into different tissue types (e.g. identify which parts of a knee on MRI are cartilage, bone, ligament, tendon, etc. We use those algorithms to help design knee replacement implants
We were apart of several large international studies investigating if COVID or previous COVID affects outcomes following all surgeries, including orthopedic surgery
We investigated what patient and injury factors increase the chance that patients will use opioids for longer than expected after orthopedic trauma surgery
Old Projects
MRIs allow for a comprehensive 3D visualization of a patient’s knee. We used a novel machine learning tool which automatically tells you what tissue type each pixel on an MRI represents to conduct automatic analyses on more than 6,000 knee MRIs
Increasingly more people are using sites like US News and Healthgrades to pick a hospital, and sites like Yelp and Google to pick a doctor. Our research found that these sites disagree with each other often and are biased over different hospital/physician characteristics.
Cartilage based tumors can be benign or malignant. Differentiating between the two often requires invasive tests. We found that our innovative machine learning technique can help diagnose malignant cancers from simple imaging like X-Rays.