Smart mouthguards: The future of head injury management?

Concussion and head injuries are consistantly at the forefront of player welfare discussions. But what technology is out there to help medics and clinicans make the best decision possible when it comes to short term and long term health of players? Instrumented mouthguards, or smart moutguards, have been around for a while but how accurate are they and are they the future of head injury management? We spoke to Greg Tierney who wrote a feasibility study looking at the different technology options and asked him six questions to help us disipher the results.

There is currently more research going on investigating each of these mouthguards which will be released in the coming months.

What information can instrumented mouthguards provide us with, and why is it so important?

For years in the automotive industry, researchers have been trying to link head kinematics to brain injury risk. The issue in sport is that we’ve never really had a valid and reliable way to do this, as sensors embedded in headgear and skin patches suffer from soft tissue artefacts, that is, the sensor / device moves relative to the skull.

Instrumented mouthguards (iMG) have embedded sensors to measure linear and angular kinematics of the head, e.g., linear acceleration and angular velocity.

Mouthguards have the potential to provide a rigid connection to the skull through the upper dentition, making it possible to measure head kinematics accurately on the field.

Furthermore, we can use these kinematics to run computer models of the head (known as finite element head models) and predict how the brain responds to these impacts. This has a lot of potential for monitoring player head acceleration exposure and aiding the sideline decision making process for player removal for concussion assessments.

How accurate is the data they generate?

There are three elements at play here. You want to know how accurately these devices are when measuring head kinematics. The only way we can do this at present is in a controlled environment, such as an impact lab. Most iMG do pretty well here. Those idealized test conditions are like a basic verification that the device hardware and firmware are functioning as they should. Essentially, the iMG system can measure the head kinematics, filter the data and transform the linear kinematics from the mouthguard to the head’s centre of gravity.

Next you want to see how the iMG performs on the field. For this, you want to make sure the iMG records “true” head impacts. You don’t want the device including a bunch of false positives or, conversely, missing true events (false negatives). Again, most devices do pretty well here, some better than others.

Last, you want to make sure that the devices are reporting accurate head kinematics from the field. You have sensors in a mouthguard, placed on your upper dentition. It’s the best we’ve got at the moment for measuring head kinematics, but it’s definitely not perfect.

There are plenty of ways artefacts can get into our kinematic signal. Artefacts are when the sensor records something that you don’t want it to. This could be movement of the mouthguard relative to the teeth, direct contact to the mouthguard during a head impact, impact to the mouthguard from the lower dentition or even biting. Any artefact in the signal can totally throw off the kinematics, so we need clever signal processing techniques to remove all of these.

Some companies have algorithms for dealing with this, but it’s often “protected intellectual property” or subsumed under “AI” or “machine learning.”

Although it’s likely better than having no algorithm at all, it can hinder cross-study comparison when different iMG systems use different ‘black box’ algorithms in their data processing pipeline. I find this is the case with many sports technologies. At present, we’re trying to develop a standardised approach for processing iMG kinematic signals based on mechanical principles that hopefully all companies could adopt.

Who are the big players on the market and how do their solutions differ?

There are a few good companies out there, four of which were involved in our validation / feasibility study: Biocore-FRI, HitIQ, Orb Innovations and Prevent Biometrics.

Each will have their own use case for their device and software platform for practitioners and researchers to use. Overall, though, in terms of the biomechanics, I believe the products aim to do the same thing in terms of measuring head kinematics accurately and reliably on the field.

Most companies will tell you they’re the best thing since sliced bread, most accurate, easiest to use, etc. We decided to design our own validity and feasibility study to determine which device suited our research needs.

How did each company fare in that feasibility study?

To somewhat paraphrase the conclusion of our BJSM paper, when compared to test dummy head form reference measurements (lab based validity), Biocore-FRI, HitIQ, Prevent Biometrics had lower and ORB had higher measurement error.

For on-field validity, Prevent Biometrics had the best false negative performance and second highest false positive performance. Biocore-FRI and HitIQ had greater false positive performance than false negative performance, and ORB had greater false negative performance than false positive performance.

One thing that stood out to me was the difference in kinematics the different iMG systems reported. I think it illustrated the need for a standardised approach to processing iMG kinematics signals. Otherwise, we’ll never be able to do cross study comparisons or compare between sports or teams that use different devices.

In terms of fit, comfort and function, HitIQ and Prevent Biometrics scored similarly high ratings. Practitioners indicated the feasibility of iMG systems to be greatest for Prevent Biometrics, followed by HitIQ, then Biocore-FRI.

We ended picking Prevent Biometrics for our studies with the Rugby Football League and World Rugby.

What is next for this technology?

I should say we’re still a bit off from this, but it’s certainly the aim of our research. What iMG mean for sport is that we now have the potential to monitor player head acceleration exposure and understand how this differs between different cohorts such as men and women, level of play and playing position. So you could imagine we get data for an out-half in rugby and see that their exposure this week or for the last few weeks was considerably above the norm for their playing position. Should we be looking to rest or reduce contact for that player this week? If we see a player who is regularly going above the norm for their playing position, should we begin to look at their tackle technique or style of play and then try to improve this?

Last, detecting whether a player has sustained a concussion on the field is challenging due to the dynamic nature of team sports. iMG could one day aid the sideline decision making process for player removal.

You could almost imagine a big impact occurs, the iMG measures it and automatically runs a rapid finite element brain model simulation. It then sends an alert to the sideline medical staff, who review the impact event and make a decision on whether the athlete can or should continue playing based on the data.

What are the biggest mistakes you see young practitioners or researchers make, and what advice would you give them to help?

The big mistake I see is allowing companies to hoodwink you into purchasing flashy new products and systems. Always look for independent validity in the literature. Don’t trust “black box” products that companies try to sell you. Try to understand the fundamental and mechanical principles that underpin how these technologies work or reach out to someone who does. You’ll hopefully then be able to see through the nonsense.

Another mistake is large, aimless data collection and mining tasks without considering the rationale or the mechanisms of performance or injury. Rubbish in leads to rubbish out with any statistical or machine learning model. Focus more on gaining good quality data and having a strong rationale for doing so, rather than just collecting lots of it and hoping for the best.

Last, we can sometimes be so proud of our R/Python/MATLAB code that we miss a simple mistake in it. Always scrutinise and sense-check your code, data and outputs.