- “Great product. But how accurate is it?”
- Testing the accuracy in-house against our robots
- Meeting Univ. Prof. Dr. Tobias Winkler, Principal Investigator – Advanced Musculoskeletal Therapies at Charite University Hospital
- Less than 4% deviation from the Gold Standard
- Preparing the first post-surgery research trial with Charite
“Great product. But how accurate is it?”
We’ve heard the question being asked hundreds, maybe thousands of times during our meetings with orthopedic doctors, therapists or other stakeholders of the orthopedic and rehabilitation industry.
In 2015 when we started re.flex we had this idea of building an industry standard for home exercise prescription. The reason behind this was to empower the patients to do their recovery at home and allow physiotherapists to automate some of the routine tasks they are struggling with, regarding customizing the exercises for patients.
We carefully studied our direct competition – products using motion tracking sensors or mobile applications with video exercises. After interacting with the tools and talking with some of their users we learned a lot from their challenges. Neither one of our competitors wasn’t able to touch that thin line between generating positive results and having an accurate, easy to use and cost-effective solution.
Mobile applications with video exercises offer a great option to complement the in-clinical sessions, but they are not able to replace the therapy, if needed. These kinds of prescriptions were often free of charge. On the other side, products which were theoretically able to replace the in-clinical session didn’t have their technology ready for this use-case.
Sensor solutions which were destined for home exercises were only able to provide a two-dimensional (2D) detection of the exercises.
This means they were tracking only one joint at a time and with minimalistic assistance based on only one metric, like for example, knee flexion.
For an efficient home exercise treatment, a 3D visualization and detection was required and we’ve envisioned this as a new standard among the digital assistive solutions for physical therapy.
In 2016 we finished the prototype of re.flex, and achieved a very good accuracy. But nobody really knew how accurate re.flex was, because everybody tested it in their own way.
Therapists and doctors were very often confronting the recordings with the goniometer, while patients were trying to stress-test the system doing repetitions with high speed to find it’s flaws.
So the question was still repeating. “How accurate is it?”
Testing the accuracy in-house against our robots
Our technical team knew that re.flex is super accurate, but they had to do complex tests and analyze large quantities of data. And even so, the “perfect movement” was something which was missing for an effective comparison. This is why we had to build two testing robots.
We planned on solving two of the biggest problems. One that all the sensors we produce fit our quality requirements and ensure the accuracy of the motion tracking.
And second that all the changes we made in the motion detection algorithm can be easily tested. Because of the physical effort and the technical knowledge required for team members to record the relevant information needed, this process was time consuming and susceptible to errors.
- You can watch a short video here.
Fine tuning the motion tracking algorithm was a laborious task which spreaded across many years as a technical focus. Bit by bit we’ve improved every aspect of the detection algorithm, which is capable of collecting 12 relevant metrics of the knee motion and assisting patients during the execution of the exercises.
But the challenge was more complex than we first believed. The size of the patient’s leg, the positioning on the sensors and the different anatomical differences we’ve observed after working directly with patients required an advanced compensation algorithm.
This algorithm was meant to compensate for the movement of the muscles when patients are exercising and clean the adherence from other influencing factors.
Meeting Univ. Prof. Dr. Tobias Winkler, Principal Investigator – Advanced Musculoskeletal Therapies at Charite University Hospital
In order to achieve the higher standards in terms of accuracy, we needed to compare our results with the most accurate solutions on the market. So we started out looking for the most advanced motion tracking laboratories in Europe. This is how we met Dr. Tobias Winkler.
Dr. Winkler is a leading orthopedic doctor and researcher. He works at the Center for Musculoskeletal Surgery, Julius Wolff Institute in Berlin. And has a deep understanding of human motion, after being involved in 90 publications focused on knee surgery, rehabilitation and accurate motion tracking.
After reading through his findings we understood we shared some common goals. Which mixes standardized exercise protocols with artificial intelligence, making it capable of automatic exercise customizations.
Dr. Tobias Winkler was impressed by what he saw and decided to join our team.
“re.flex it’s a solution where the patients can train this themselves. And they can adhere to all the videos and suggestions that are in the system. They can do that everywhere without needing to go to the studio, the physiotherapy centre or to the clinics, which of course in general costs. The hope is that the doctor or the physiotherapist can then also control the progress of the training by just looking into the data of the patients that is sent to the respective doctor or physiotherapist”
Later on Dr. Winkler introduced us to Dr. Phillip Damm, the lead Biomechanical engineer of the Orthoload Lab, at Charite. Located in Berlin, Charite is the largest University hospital, and the motion tracking department is very large and innovative. It has the largest VICON implementation, with 20 motion tracking cameras recording every aspect of the movement, on the entire body.
Achieving a high level of accuracy was the most important metric for re.flex. A high accuracy builds enough trust to allow orthopedic doctors and therapists to prescribe patients with home rehabilitation after surgery. And of course, for patients to accept working out with pain. Alone. At home. This is not an easy thing to achieve. Having an accurate representation of the real-time movement and a good understanding of the target of the exercise was essential for a safe rehabilitation at home, assisted by the sensors.
Less than 4% deviation from the Gold Standard
In 2020 we managed to complete the Charite University accuracy study.
Our innovative solution scored comparable results with the Gold Standard, the VICON system.
The measurements were done using different types of exercises.
We tested the supine exercises which were mostly used in post-surgery rehabilitation. Then we continued with the standing exercises, for the second part of the rehabilitation or for prevention.
Analyzing the data we discovered less than 4% deviation from the standard. Which is a very good result if you put all things in the right perspective.
The VICON-Nexus system costs in a similar installation as the one at Charite are anywhere between 1 and $1.5 million. The system features up to 24 motion cameras, with optical markers within an accuracy of up to 0.3 mm. The markers need to be placed in a certain position, so patients are not able to operate the system alone. And it will always be a clinical device.
On the other hand, re.flex’s main use-cases are for patients exercising at home. Results from the Charite study were needed in order to validate our readiness to explore complex use-cases as total knee and hip replacement rehabilitation or ACL rehabilitation.
re.flex’s Chief Technical Officer and Co-Founder Andrei Kluger was satisfied with the results.
“The Charite validation with Vicon was a big milestone for me as a CTO of re.flex as it proved that our solution meets the most rigorous standards of the industry and our results can be independently validated.
Vicon is an industry standard for 3D motion capture and its precision has been thoroughly validated in research and industrial facilities”
Andrei Kluger, CTO & Co-Founder @ re.flex
re.flex’s 4% deviation from the standard of motion tracking is a very good result, considering the theoretical accuracy of the IMU sensors is 2%.
“The data we’ve recorded have on one hand validated our motion assessment models and on the other hand aided us to improve the product even further. It’s very important for each new technology to be validated with the industry standards.
We’ve gained lots of confidence after the results and it helped us with the product marketing. The Orthoload laboratory from Charite is a state of the art lab used for research purposes and it showed this from the star”
Preparing the first post-surgery research trial with Charite
Using the results of the Charite study, re.flex is set to add the last touches on the compensation algorithm, used in eliminating the muscle activity from the measurement. After this we planned to do a series of improvements in the post-surgery product, one which already gained traction among the orthopedic surgeons.
In the background the a team from re.flex and Sporlastic is preparing a post-surgery study, which will be coordinated by a team from Charite University Hospital and Dr. Tobias Jung.
This article wasn’t for you? Check our Youtube channel!