I believe certain visuals tend to help with storytelling. For this post, I have a video to help better understand my narrative.
About a couple of years back I spoke to a small VR startup about using AI to help with fitness and health. And one of the uses for video analytics is to track human movement (or pose detection as it’s called in the industry). Instead of looking at use cases in athletic performance enhancement for sports, there are use cases for medical rehabilitation that can help with the overall quality of life, and maybe even cost reduction due to automation, and better patient-to-specialist ratio. What am I talking about specifically?
As you can see in the video, there is an elderly man who is going through a rehab exercise with a physical therapist. I was told that the man suffered from a stroke recently and was slowly recovering. He was in a program to study if the use of digital technology can further accelerate the healing process vs conventional methods. You can see that he is wearing a VR headset to play a specific set of exercises that mimics daily activity. And using Computer Vision, we super-imposed a ‘pose wireframe’ on the man (and the therapist, but she’s not being measured)
I did this side project for a friend who was part of a team of experts looking into movement rehabilitation for stroke victims and people with disability. He explained one of the key issues in Malaysia is the lack of specialists to consult on all the patients at regular periods. And in many cases, patients live too far away from any nearby specialist centre to be able to meet regularly for rehab. The idea I was trying to prove for the friend was, can we use AI as part of his program to remotely monitor and support the patient to do rehabilitation exercises and can it be done at a reasonable cost?
As you can see in the video, I built a simple application that can detect humans in the video, and apply a posable frame on the human. Each of the joints of the body can be measured as data points in terms of relative position and angles. These are good enough data points to understand certain measures of human movement and to start creating a solution to help determine how well is the patient moving in relation to a normal healthy human, and measure against his previous sessions over time.
On top of that, I had proven to the team that it can be affordable and the AI can run on an affordable modern desktop computer, with real-time tracking capability. This will allow units that they want to deploy to be made available in many places around the country, especially remote and rural areas. And it will give rehab specialists more time to monitor the patient’s progress, and also potentially have better quality data for assessment.
Suffice it to say it is enough evidence for the team to start building out their program to add the AI solution. And the benefits are very clear in terms of health, rehabilitation, life expectancy and even cost. I’m happy I could contribute to something positive for the community.
