Will the Robots Save Ultrasound? Save

Ultrasound has a lot going for it in rheumatoid arthritis and psoriatic arthritis. It is free of ionizing radiation, yet can deliver helpful insights to the bedside, educating our patients along the way. As we seek to prognosticate more and delineate better, ultrasound should be our friend.

That is not news. It was emphasised strongly at this year’s ACR Convergence in the proposed ACR Guidance for Use of Musculoskeletal Ultrasound in Rheumatoid and Psoriatic Arthritis (16S06), but also ultrasound training has started to permeate many rheumatology training programs globally, and is incorporated into many rheumatologists’ modern everyday practice.

Despite all its benefits, however, ultrasound’s daily clinical use remains the domain of only a subset of interested rheumatologists. If the benefits are so clear, then why is it not universal? A few key provisos have stymied its uptake. Performance remains operator-dependent - and even if you have been taught by the best, unless you continue to do MSK ultrasound regularly and at scale, the full benefits may prove ever more elusive. Additionally, even limited examinations also take substantial time, and in an era of rheumatology workforce shortage, lingering questions remain as to how rheumatologists’ time can be best harnessed.

So this is not an article about ultrasound scepticism, but about ultrasound realism. It is in everyone’s interest to find a practical way to harness the benefits of MSK ultrasound in inflammatory arthritis, on scale.

Perhaps the robots will save us?

In the late-breaking oral abstracts at ACR Convergence this year, the final abstract presentation was from the Danish rheumatologists behind ARTHUR and DIANA: the robot which holds the MSK ultrasound probe, and the artificial intelligence brains which empowers it (abstract L20).

The setup is simple. The patient approaches the robot, introduces themselves, follows the instructions (including applying their own gel), and then, reminiscent of an automated car wash, the machine scans their hands. It takes 15-20 minutes for both hands, and no clinician time. The interpretation is automatically uploaded to the patient record, ready for the treating rheumatologist to review.

In the presented study, the robot repeated the scan twice, looking for grayscale synovitis and Doppler activity, and each of those assessments was compared to an expert rheumatologist, to see who could pick up the most disease activity. In this, the robot was quite consistent in its interpretation (with a repeatability binary agreement of 88-90%), but more critically, the machine did just as well as the expert rheumatologist in detecting relevant findings. As a whole, the robot actually made a correct patient-level assessment more frequently than the expert rheumatologist.

This is not to say that expert humans don’t have their advantages. Expert rheumatologists can take on targeted examinations based on history, although even a targeted exam by a rheumatologist still takes up more rheumatologist time than when a rheumatologist is not doing the scan at all. The scanning protocol in this study is highly stereotyped, but often enormous value with ultrasound comes from scanning beyond the small joints of the hand, particularly from tendons and entheses. The robot can’t inject at the same time, yet, nor can it educate the patient while it scans - although those seem like solvable problems in due course. And cost and accessibility may yet be problematic for these highly specialised robots, too. Nevertheless, it seems possible that ARTHUR and DIANA might help to broaden rheumatology ultrasound workforce capacity, and even free up expert rheumatologists to do more challenging and novel scanning.

Data on overall performance (including false positives) will come soon, and will be key to how much we as clinicians accept these results. However, patients have less qualms than we might imagine - previously, the authors have reported that patients actually quite like the robots.

In the meantime, those of us who find ourselves stuck in the backwaters of ultrasound reality can dream of future help from an army of robot friends.