Finding Normal

Wired magazine ran an article 'Finding Normal' in May 2008. Diagnostic techniques like MRI or CT are becoming tools of research for predictive medicine and any apparent smudge in images could be interpreted as a lesion, which, at surgery, is found to be benign. Mr. Goetz criticizes the field as a whole, all the way from the NIH down to the individual scientist.


It's tempting to believe that scientists pursue research this way: 'Eureka! A spot on MRI! Nobel, here I come!'. In general, scientists don't do this and most realize that their grant could never be funded this way.

The NIH does give priority scores to indicate a project's novelty, originality and scientific merit, but as any study group at the NIH will tell you, understanding what's normal is a key component to whether the research is funded. Normal, in this case, is the control group - and characterizing this group of people defines what's abnormal.

As an example, Dr. Subramanian hypothesizes that atherosclerotic plaques have higher tau-score on MRI. Dr. Subramanian has investigated tau-score in vitro and realizes that it correlates well with atherosclerotic fat content and suspects that high tau-score can be used to characterize plaque risk.

Should Dr. Subramanian...

A. Quantify tau-score among 40 patients with heart disease?
B. Correlate tau-score with the gold standard in patients with heart disease?
C. Quantify tau-score among 40 healthy normal subjects.

A, B and C are all necessary to fully characterize the disease and tau-score. In fact, scientists look at a healthy population all the time to estimate the sensitivity and specificity of the diagnostic technique. How specific and sensitive is the technique, if 40% of healthy normal subjects have an atherosclerotic plaque with high tau-score and apparently high fat content? Specificity and sensitivity scoring is ubiquitous to MRI and science in general.

It is true that we have actually surpassed our capacity to interpret the results of diagnostic imaging. This doesn't preclude scientists from doing good research, but suggests only that more scientists should use imaging as a tool.

Predictably T2

The white bread of MR imaging uses T2 to generate contrast between healthy and diseased tissues. What a surprise when boring, cookie-cutter T2, so overused and under-appreciated had a big boost in interest at the cardiac MR study session this year at ISMRM.

In a room that fit 500 people, about 20 (5000+ are at the ISMRM) made it to the early evening discussion led by Dr. Andrew Arai at the National Institutes of Health about a unique application for T2-weighted imaging. Buried under the glamour of delayed gadolinium enhanced MR, full-body magnetic resonance angiography and cine, T2-weighted imaging is the only one that apparently gives useful contrast for acute myocarditis. In this case, the infected area becomes brighter on MRI, likely because of edema and inflammation. A short discussion on the research appeared in Circulation recently.

This is especially nice because T2 contrast is endogeneous. It also doesn't require specialized hardware or pulse sequences to use. You could potentially quantify the tissue to monitor the muscle longitudinally or run a multisite/multivendor trial with reproducible results. The image to the right shows the inflammed tissue as bright signal on T2 MRI in a comprehensive study of several techniques.

Clearly, this is also an opportunity to use single-shot T1R, which would fit nicely in the RR interval of the cardiac cycle. It has improved dynamic range, reduced diffusion and would reduce any effect of cardiac muscle orientation on quantification. Also, anyone with a complicated T2-weighted sequence (T2-prepared bSSFP, variable flip angle fast spin echo, etc) looking for a useful non-brain application could find one.