Inspiration
As we watched our parents and grandparents get old and have to go to physiotherapy, we realized many issues in the industry as a whole. Doctors and Physiotherapists primarily diagnose using visual observation and patient-reported pain. One of the best ways to aid doctors in diagnosing patients and helping them fix their pain quickly is to use 3d motion capture. Yet the only options to use this type of EMG diagnosis can cost roughly 12500 dollars for the lab systems and an average of 350-450 dollars per session. This leaves a blind spot for doctors who have patients who are online, or cannot afford these systems that help predict future pain.
What it does
We built a tool to help physiotherapists fill this gap in patients. Atlas takes standard smartphone video and turns in into a full OpenSim bio-mechanical simulation. We can get real joint angles, muscle forces, joint loading extremely close to what research labs compute them in a cheap, quick method. Then we provide an Intervention Engine, a powerful tool to assist doctors in diagnosing your issue quickly and methodically. It allows clinician to test how different aids such as belts, braces, or orthosis could redistribute muscle load in that patient.
Challenges we ran into
Getting from a video filmed to a actual simulation was not an easy feat. We had to give options for dual filming to improve accuracy as well as test edge cases where we had to reject unsynchronized footage that could misrepresent the actual forces. We struggled but finally were able to compile multiple datasets together to ensure that our simulation worked through all types of motions so that we could provide the most accurate information to the doctors.
Accomplishments that we're proud of
The Intervention Engine and our Long Term Motion Simulation are the 2 things we are most proud of. We can take just a video of the motion and clamp the range of motion the way a rigid knee orthosis would. We can then simulation muscle equilibrium physics recompute what happens to the remaining crossing muscles without just using a guess. Using our Motion Library we can make our Long Term Motion Simulation that can see how the model behaves on tasks outside of what was filmed. This addresses a gap allowing clinicians to see how their patients body would respond to various motions instead of having to guess or test over many sessions wasting valuable time and money.
What we learned
We learned a lot about how simulations and algorithms are going to be used in the health space. We believe as a team more firmly then ever that AI and new research and innovation should be used to help blindspots in a doctors work. Rather than attempting to replace or override a doctor's intuition, we should seek to provide tools that help their work to be faster and cheaper.
What's next for Atlas
I think we have more to build on this project. Some of our steps could include building static optimization which would further the accuracy of the patients inverse dynamic torque calculations. Our product that already shows a strong upper hand from typical methods in being unbiased to different body shapes could be furthered by taking heigh, weight and sex as inputs. Finally, we would love to have a physiotherapist come work with us and looks at our work and gives us more ideas on what would help them.
Log in or sign up for Devpost to join the conversation.