To the Editor:
We read with interest the letter in this issue by Hall-Craggs and colleagues1 discussing the developing field of diffusion-weighted imaging (DWI) and its application for quantification of inflammation in rheumatology. As proponents of quantification in medical imaging, we agree that DWI does indeed have some advantages. Its ability to objectively quantify diffusivity of water molecules in the brain was well established many years ago2 and is now absolutely essential for magnetic resonance imaging (MRI) of stroke, providing unique information about brain cell injury due to hypoxia. DWI is now often used for detection and characterization of tumors3. However, its application for assessment of inflammation outside the brain is at an earlier stage of development. One potentially exciting application may be the assessment of synovitis, where conventional sequences have difficulty distinguishing inflamed synovium from effusion without intravenous (IV) injection of a contrast agent. If DWI turns out to be a suitable substitute for the IV injection4, it would be a significant advance, especially for children.
It has been shown by Vendhan, et al that DWI measurements of sacroiliitis correlate closely with observed quantification of bone marrow edema on short-tau inversion recovery sequences (R = 0.85)5. In this publication, the measurement of apparent diffusion coefficient (ADC) required an observer to place regions of interest (ROI) on DWI images. A second observer was told how many ROI to place and on which images to place them, but not exactly where to place them. The resulting interobserver agreement for the normalized ADC measurements was 0.64. This may be satisfactory but it is not impressive for a measurement that Hall-Craggs, et al1 emphasize is objective. Of course things improve with experience, and 2 years later the same author group had intraclass correlation coefficient for DWI of the sacroiliac (SI) joints in the range of 0.81–0.98, using different methodology6. Recently, Bradbury, et al7 confirmed that DWI corresponds closely with observer quantification of bone marrow edema (BME). Their conclusion that “DWI has excellent performance as a diagnostic tool in distinguishing axSpA [axial spondyloarthritis] from noninflammatory causes of back pain” may be true but not based on their own data, because their study design required subjects in the control group to not meet the Assessment of Spondyloarthritis international Society (ASAS) imaging criteria. In addition, the mean Spondyloarthritis Research Consortium of Canada SI joint score for this group was 0. In other words, their control group did not represent a typical low back pain population, which frequently demonstrates small foci of BME in the SI joints8,9,10. More recently, it was shown that DWI may improve the specificity of a diagnostic MRI scan without significantly changing its overall diagnostic performance11. In this publication, the area under the curve for diagnostic ascertainment using measured ADC values was inferior to subjective assessment of the MRI, although the difference was not significant. This is not surprising, because regardless of whether quantification of BME is helpful for diagnostic purposes, it does not incorporate information arising from localization of BME and the contextual interpretation with other imaging findings that contribute enormously to diagnosis such as erosion. It is problematic that many reports refer to a “positive” MRI, as defined by the ASAS criterion, as now being “an important component of diagnostic pathways in spondyloarthritis12.” This criterion, which is based solely on the assessment of BME, was generated by consensus for the purposes of classifying patients with axSpA and was never intended to be adopted for diagnostic purposes.
A whole range of new MRI mapping techniques are now available permitting quantification such as T1, T1ρ, T2, and T2* mapping as well as DWI and chemical-shift encoded MRI. It seems likely that some form of quantification will become a significant part of many future applications of MRI. In the meantime, it should be emphasized that nearly all these MRI techniques still rely in part on medical imaging and expert interpretation of both the imaging and numerical results. MRI “maps” are in fact “images” within which the varying color/brightness of the pixels represent a particular property of the tissue. An ADC map (image) is objectively created but still requires interpretation, and when an observer places a region of interest over an area of the map, the results may be subject to significant observer variation11. It is hoped that artificial intelligence and computer-aided detection will have a major role in assisting the human observer to better interpret MRI results in the future, thus easing the load for the observers and enhancing patient care and clinical trial research. While objectivity is preferable to subjectivity, objectivity alone does not confer usefulness. There are pathophysiological reasons why the application of DWI for diagnostic ascertainment of bone marrow inflammation in rheumatology may have limited clinical utility, and its future use is likely to primarily focus on clinical trial research.
REFERENCES
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.
- 12.