About Agile Kinetic
Agile Kinetic is a tech-centric business striving to use technology to improve societal health. Its platform uses AI to monitor patients’ musculoskeletal health, which in turn improves information-sharing between patients and their clinical teams and informs the decisions and recommendations made by them.
Pose Estimation takes an image or video of a person from any device and automatically plots their joint positions. We were initially looking at using it to automate coaching and sports, but then we met a surgeon who thought it could be really useful in their sector.
We started developing a solution for physiotherapists and surgeons
The aim of it was to allow them to be able to monitor their patients following an injury or surgery. This is because it’s important for them to know whether a patient's range of motion is improving or not over time.
We’d built a prototype app-based platform for patients to access
Patients could enter things like their daily comfort score, for the app to monitor. They could also upload requested images of their joints when flexed or extended, which would then record a patient’s range of motion via the Pose Estimation technology.
We hit upon two challenges that showed us that we needed to do further R&D
One challenge was that we needed to find a way to motivate patients to use and engage with the app; we needed them to submit daily information regarding their pain and comfort scores and take on the recommendations that the clinicians fed back to them, based on their range of motion and comfort levels.
The other challenge was that we needed to find a way to help patients take the required images of themselves. All we needed to begin with was two images, one of the joint fully flexed and one of the joint fully extended. We were developing the app for older people who weren’t digital natives. They needed to be able to navigate the app and stand in the right positions while capturing the images. It became apparent that we’d need to develop a way to coach them through taking those images.
Clwstwr funding gave us the chance to explore options through R&D
We split the project into two R&D areas – 1. Finding a coaching method that would guide patients through getting into the required positions and taking the images, and 2. developing a way of using gamification to encourage use.
We experimented with three ways of coaching patients to take the required photos
Option 1 was to have an automated trigger on the camera. This would allow the patient to be in position, at a distance from their phone, copying a person on the screen who’s doing the required movements. Meanwhile, their phone camera would automatically take an image of them at certain times to record their range of motion.
Option 2 was to have a voice-activated trigger. It’d allow the patent to move away from their smartphone, get into what they felt was the right position, then use their voice to activate the trigger.
Option 3 was to get patients to film themselves performing the whole sequence of movements, then we would have to find a way, at the back-end, to get to the right points in the movement to then record them.
We wanted to do in-person user testing with 40 National Exercise Referral Scheme participants
However, the pandemic forced us to change our plans. Instead, we used a slightly smaller cohort and did the testing over Zoom. We figured out a way to share our screen with them in order to give them the instructions, while recording them at the same time so we could measure their range of motion. Then, we gathered their feedback on the different options.
What we thought would be the most user-friendly approach wasn’t the most popular
We thought users would prefer option 3 because it’d involve the least amount of effort on their part, but it was actually the least accurate and least preferred option. The reason being was that our users prefer to be guided really clearly; they’d often freeze or move incorrectly because there wasn’t clear guidance. The other methods, such as the automated trigger, proved to be the most popular because we gave more guidance to patients as we needed them to be in the right position when they took the picture. This all showed us how important user friendliness and thorough guidance are to our users.
We also came up with a hardware solution to reduce photography errors
The basic task of getting somebody to take the camera and get it to face them with their whole body in focus was a big challenge that we didn't really expect. So, we experimented with lots of different low cost tripod setups to see if there was a mass market solution that’d hold the phone in the right position. We ended up sourcing some really cheap and simple solutions involving a clip. With PDR, we designed a modified version of it to create a prototype, coming up with a design that we could produce for less than £1.
By combining elements of both outcomes, we now know how to get people using the technology and we’re able to provide them with the equipment they need.
The gamification research was really interesting to us
We recruited a graduate in game design to do secondary research for us for about two months, with the task of trying to find out what kinds of things motivate older users and non-digital natives, and what might keep them engaged with the app. They came up with a huge amount of research and suggestions, which we then took forward by way of prototyping some of the most promising options.
The most engaging solution was something that could easily seem trivial at first
Our users were most engaged when they were being rewarded with a congratulatory message and haptic feedback (such as the phone vibrating and making a happy noise for a brief moment) after completing certain activities.
In fact, the feedback was that we needed to go further with this kind of reward; instead of getting it at the end of completing all of a day’s exercises, they wanted it every time they completed each set of exercises. That way, they’d get rewarded more often and build streaks. This is the next thing we’d like to implement.
We’re really happy with the progress we made and have already done follow-up R&D
Since finishing our Clwstwr project, we’ve been building on our underlying technology with a machine learning researcher and looking at validating our software to prove its accuracy in taking measurements. We just closed a round of investment involving a medical company that’ll be joining our board, a big milestone for us. Our next milestone is getting the required UKCA marking to take this to market.