Lessons emerge from 20 years of clinical MRI

MRI began to be used as a clinical tool during the 1980s. At that stage, it was extremely expensive and very slow, with individual procedures taking around an hour. It was also noisy and very conducive to claustrophobia. Because examinations took so long, the images were prone to motion artifacts.

MR systems were regarded at this time as "toys" for musculoskeletal radiologists and neuroradiologists. Twenty years on, this view persists in many countries and radiology departments, making prompt access to modern MRI extremely difficult. This despite the fact that MRI can be used for many of the same applications as CT, but without the ionizing radiation.

In the early days of MRI, each individual sequence took a long time. But how often is it that the first sequence is the most appropriate sequence, the one that gives the diagnosis? This aspect was investigated at Addenbrookes Hospital in Cambridge, U.K., leading to the discovery that beyond two sequences, you often don't make much more diagnostic progress (Coulden RA, Dixon AK, Freer CE, et al. Clin Radiol 1991;44[6]:393-396.).

This problem remains pertinent today. Consider the example of liver MRI. One examination may take 40 minutes and include 16 or 17 sequences with varying fat suppression. How often does that extra sequence actually change the diagnosis?

Even if you plan to perform more than one MRI sequence, you should always perform the most useful sequence first. After all, this may be the only one that you get if the patient chooses to climb out of the magnet bore. What counts as the optimal sequence may change as technology develops. Most radiologists would once have agreed that you need axial T1-weighted imaging for lumbar spine MRI. You can now see the individual nerve roots, thecal sac, and disk material on T2-weighted images, making the axial T1-weighted sequence redundant.

At least one sequence with a large field-of-view should, ideally, be acquired for most applications. The smaller the field-of-view, the better the image quality, but there can still be a lot of useful diagnostic information on the initial localizer sequence. With a hip MRI examination, for instance, one sequence should probably show the full pelvis and both hips.

With MRI, it is essential to know your anatomy inside out. In fact, as always with imaging, you should have a better knowledge of anatomy than your clinical colleagues. You also need to recognize artifacts, know what is causing them, and learn how to avoid them.

It is important to recognize situations in which errors are prone to occur. Radiologists who audit their own work and work in teams are less likely to make significant mistakes than lone workers. Many radiology departments in Europe have a policy of double-reading mammograms; should we also be double-reading MRI?

It has been proposed that lumbar spine x-rays could be replaced by a single sagittal T2-weighted MRI sequence. This use of MRI as a generalized screening technique will require a complete change in mindset. MRI has got to get quicker, and this means doing fewer sequences. If you did just one standard T2-weighted sequence for the lumbar spine, you could see a patient every seven minutes and run through as many as 70 patients a day on some machines. Those requiring more detailed imaging could then be called back for a longer examination.

MRI is still regarded as too expensive to be used for screening. It should not be. In most countries worldwide, it is standard practice to x-ray suspected scaphoid fractures. But this will not actually tell you whether there is a fracture or not. Even scintigraphy may be negative initially. Why not go straight to MRI? You would then see the edema and any fracture line.

As MRI continues to evolve, it is vital to keep reading. Radiologists will not be doing in 20 years time what they are being taught today.

This column is based on a lecture by Prof. Adrian Dixon, a professor of radiology at the University of Cambridge, U.K., given to the Radiology Trainees Forum at the 2007 European Congress of Radiology.