Can We Rely on the Interpretation of MRI Scans?
MRI scanners are a wonderful invention and continue to advance, giving us finer and finer pictures of our anatomy. As is typical with all fast-improving technology, the human aspects, the culture, struggle to keep up.
An MRI scan means nothing without an interpreter. It’s a language most of us don’t speak. Ordering a scan presupposes that there is someone to say what it shows and crucially, what it means for the patient. Someone has to put the patient’s story, the examination and the scan together and form a conclusion about what the next step is.
The developments in machine learning mean that reading x-rays and MRI scans will soon be done by machines. The consistency of analysis this promises will still not be enough if there is no one to sit down with the patient and discuss the results.
But this post is about the consistency of analysis of MRI scans. The ‘someone’ in these cases is a radiologist. Radiologists are highly trained doctors who specialise in imaging and in various types of interventions where imaging is used.
Since radiologists are human we should expect variation in the way MRI scans are interpreted. But how much?
A 63-year old woman with a history of low back pain and right leg pain in a particular pattern was sent to 10 different imaging centres and given MRI scans for her problem. The centres were unaware that she was being scanned anywhere else.
The researchers looked at what the radiologists reported each time for the same scans and the results were very illuminating.
In the 10 reports they found 49 distinct findings reported for the scan (disc protrusion, facet joint arthritis etc). 16 of those findings only occurred once across the ten scans, in other words the other nine radiologists did not report that finding. None of the 49 findings occurred in every scan report, making us wonder about the validity of the findings.
Highly skilled and highly trained radiologists were doing the best job they could in reporting on these scans. But the results were so variable and inconsistent that it calls the process into question. How can we make important decisions about a patient’s medical intervention if the scan finding are unreliable?
The scan looks so solid, on the screen in black and grey and white. It’s hard to accept that we might all see very different things and come to very different conclusions about what to do next for a person in pain.
And that’s without even considering the matter of what the scan results mean for our future, should they be valid in the first place. That’s next.
Herzog, R., et al. (2017) Variability in diagnostic error rates of 10 MRI centers performing lumbar spine MRI examinations on the same patient within a 3-week period. The Spine Journal, 17(4), 554-561.