Half of people aged 60 or more experience shoulder damage, mainly because of wear-and-tear. Treatment is usually surgery followed by a lengthy period of physiotherapy – short sessions of careful physical manipulation that avoids injuring recovering muscles and tendons. But research shows that shoulders recover better and more quickly if they are manipulated more often and put through a wider range of movements. This demands a better insight into the working of the shoulder joint, which is where robots can play an important role. In a one-year Cohesion project, Dr Ajay Seth at Biomechanical Engineering (BME) collaborated with Dr Luka Peternel at Cognitive Robotics (CoR) and post-doctoral researcher, Dr Micah Prendergast to design a ‘biomechanics aware’ robotic system that delivers optimal rehabilitation of rotator-cuff injuries within self-defined safety limits.
Where did the idea for this Cohesion project come from? “Before I came to Delft, I had developed a model to look at the mechanics of the shoulder and what muscles are doing to generate certain movements,” explains Seth, whose background in Computational Biomechanics enables him to create musculoskeletal models to understand better how humans move.
“And I had already used your biomechanical models while working on human-robot collaboration,” adds Peternel. “Understanding the ergonomics of human co-workers enables robots to adjust their way of working with humans to make it easier for humans to work with robots! So having already used Ajay’s models, I thought our areas of expertise would be a perfect combination for a Cohesion project.”
Within the BME department, experts in shoulder anatomy and modelling pointed to rotator-cuff muscles, which are highly prone to injury: in fact it’s estimated that about 22% of the general population will experience rotator-cuff damage at some stage in their life. “This is because the shoulder joint is a complex structure with a wide range of movement,” explains Seth. “There’s very little bone there and it’s mostly held together by soft tissues and muscles, including the four rotator-cuff muscles.”
If a rotator-cuff muscle or tendon is damaged, it is typically repaired surgically - then follows a long period of physical therapy. But while the literature shows that shoulders recover better and faster the more they are exercised, physiotherapists have to be extremely careful not to overdo either the range or the amount of manipulation because of a relatively high risk of damaging the recovering soft-tissue. Part of the problem is that it is difficult to see what’s actually happening deep inside a person’s shoulder and therefore determine the degree of possible movement in each direction without risk of re-injury – all of which means that post-operative shoulder rehabilitation tends to be overly conservative.
So having already used Ajay’s models, I thought our areas of expertise would be a perfect combination for a Cohesion project.
Post-doctoral researcher Prendergast, whose expertise is in Mechatronics and Medical robotics, joined the team in 2020: “We needed to understand from the biomechanical side exactly what was happening inside the patient’s shoulder when they move their arm around so the first thing we did was build strain maps for the tendons of the rotator-cuff muscles.” Strain maps are spatial maps based on movements in three dimensions so if your arm is in one particular position, the map reveals the strains that injured tendons are experiencing at that point. “Our primary goal was to use these strain maps to inform the robot of the tendon strain in each of the four rotator-cuff muscles throughout the range of motions.” Having done this, Prendergast was able to plan trajectories that avoid putting undue strain on these tendons: “We created a map of possible arm movements that the robot can use to plan a course of safe movements. So essentially we can tell the robot to manoeuvre around this particular pose of the arm because we know that if we go in there, we are straining one or more of the rotator-cuff tendons.”
“It’s a bit like having a ship and a chart of the reefs - you know where you have to go to avoid the reefs,” adds Peternel.
Being able to chart the space of all possible movements and know where both the safe as well as the unsafe movements are, it should now be possible to create a new range of exercises. Furthermore, using the images from each patient’s CT or MRI scan made before surgery, it should be possible to customise physical therapy to each patient and their specific injury. Prendergast: “So it’s actually a very powerful tool and the part that brings it all together and allows it to work is that it’s not a human moving the patient around, it’s a robot that can be safely controlled to implement these more complex exercises which is really where the human-robot interaction comes into play.”
Their paper ‘Biomechanics Aware Collaborative Robot System for Delivery of Safe Physical Therapy in Shoulder Rehabilitation’ (June 2021) was published in IEEE Robotics & Automation and Seth, Peternel and Prendergast have applied for a patent on their system. They also have plans to continue working together as the collaboration worked for them all: “Now I appreciate the complexity that’s behind this model,” says Peternel. “Before the Cohesion project, I was just using it but now I see how much work goes into applying the model to something as complex and advanced as a shoulder. Also we’re a good team!”
For Prendergast, the project was about learning to pull out the most important pieces of information and put them together rapidly: “We really hit the ground running and we were ready to write this paper within about six months so we moved quickly.”
“And for me as a modeller,” adds Seth, “it’s always been about how I can use the models to make people’s lives better so robotics is the perfect answer – the ying to the yang; now we have this insight into the person, what can we do to improve their situation and the robot can look at it and execute a plan, closing the loop with this person.”