Is There a Role for Diffusion-weighted Imaging in Patients with Brain Tumors or Is the “Bloom off the Rose”?
================================================================================================================

* Robert D. Zimmerman

It was the day after the department's holiday party, the traditional time for our annual MR protocol meeting. We meet that day on the theory that everyone is either too mellow or hung over to argue much. Yet all eyes turned anxiously to me when someone suggested we remove diffusion-weighted imaging from our seizure protocol. After all, it was I who had whined to our MR vendor for almost 2 years that we had to have diffusion imaging capabilities immediately. When we had diffusion imaging equipment, it was I who had insisted that it be part of all our brain protocols. But at this meeting I shrugged and agreed. After a few years, it had become clear, even to a zealot like me, that diffusion-weighted imaging was not going to provide useful information in all disease processes.

There is a natural history for new diagnostic tests. The initial results are amazing, the praise overwhelming. The test is viewed as having almost magical qualities. After a while, reality sets in. The initial enthusiasm wanes as experience accumulates. This is where we are now with diffusion-weighted imaging. Many recent publications have documented the limitations of diffusion imaging. In this issue of the *AJNR,* Kono et al (page 1081) continue this trend. The authors used diffusion imaging to examine 56 patients with the three most common types of intracranial neoplasms: gliomas, metastases, and meningiomas. They found that there was much variation within each group and no significant difference in intensity among groups on either diffusion-weighted images or apparent diffusion coefficient (ADC) maps. Tumors in all three groups ranged from hyper- to hypointense. In addition, the authors could not distinguish between vasogenic edema and tumor infiltration by visual inspection of images. They concluded that diffusion and ADC images cannot be used to differentiate between tumor types or to assess tumor grade. A quick inspection of the images confirms that diffusion-weighted images and ADC maps do not provide much useful information when compared with conventional MR sequences. In some cases, the lesions are difficult to find, let alone characterize.

But all is not lost. As our knowledge of the virtues and limitations of a test increase, we develop a more mature, if demystified, appreciation of its value. This is true for diffusion-weighted imaging. There were positive findings in Kono et al's study that provide us with direction for future research and use of diffusion imaging in assessment of intracranial neoplasms.

The most important trend is toward the use of quantitative diffusion imaging techniques. When ADC was measured, there was a significant difference between grade 2 and grade 4 gliomas. High-grade gliomas had lower ADCs than did low-grade gliomas. Thus, quantification reveals significant differences that were not apparent upon direct visual (qualitative) inspection of either diffusion images or ADC maps. Assessment of ADC values has also yielded interesting results in evaluation of other process such as acute infarction and normal aging.

It has always been possible to measure the physical parameters that underlie an MR image (eg, T1, T2, magnetization transfer ratios [MTR]). However, this is a time-consuming process often requiring special image acquisition protocols and intensive postprocessing. Moreover, these measurements have not proven to be useful in routine clinical practice. There is no guarantee that things will be different with diffusion imaging, but it seems that diffusion measurements may be of more practical value than other physical parameters of MR. ADC maps are easily generated from routine fast diffusion-weighted imaging by use of software available on many MR systems. Therefore, diffusion values can be obtained without a great deal of effort. Second, ADC may be a more a direct indicator of changes in the brain than are other physical parameters, because in the majority of cases, these parameters evaluate submicroscopic, molecular-level phenomena. T1 and T2 reflect changes in the interactions between water protons, whereas MTR reflects interactions between somewhat larger “structures”, macromolecules and their hydration shells. All of these processes occur below the level of basic biological structures. The degree of diffusion, on the other hand, is strongly affected by microscopic biological structures such as the number, type, and spatial arrangement of cells. These structures create barriers to the free diffusion of water. Therefore, changes in diffusion may more directly reflect changes occurring within and between cells.

Microscopic features are, of course, precisely those used by pathologists to diagnose and differentiate neoplasms. In Kono et al's investigation, the authors evaluated the relationship between ADC values and tumor cellularity as determined by an automated quantitative evaluation of cell blocks. They demonstrated a good correlation (r = −0.77) between ADC and cellularity in gliomas. This is an encouraging finding, because cellularity is an important histologic determinant of glioma grade. The authors point out that other histologic features that are known to influence tumor grade may also contribute to ADC values, including nuclear cytoplasmic ratio, tumor matrices, and extent of fibrosis and gliosis. Until now it has been difficult to assess the relative contributions of these factors to ADC. Kono et al report on an important first step toward more precisely determining the effects of histologic variables on ADC. In the future, correlation between quantitative assessment of histologic measures and ADC values may improve our ability to assess tumor grade.

The most disappointing finding in Kono et al's article is that diffusion-weighted images, ADC maps, and ADC measurement cannot be used to determine the extent of tumor infiltration and to differentiate infiltration from peritumoral edema as previously reported (1). This is one of the “holy grails” of glioma imaging, because determination of extent of infiltration could affect both treatment and prognosis. Because infiltration occurs within and along white matter tracts, diffusion tensor imaging may yield useful information in the future. Another avenue for future research is high-b-value diffusion-weighted imaging that could allow for a more accurate assessment of infiltration as well as open a window for imaging the intracellular environment within tumors.

As we proceed to evaluate diffusion-weighted imaging of tumors, it will be important to ask useful questions. These will vary depending on the type of lesion under investigation. In Kono et al's article, the authors evaluated the three major groups of intracranial neoplasms by using the same techniques. However, within each broad group there will be different issues that need to be addressed that will require a more customized approach.

For instance, the finding that ADC of meningiomas did not correlate with histologic subtype is not a significant concern, because subtype does not determine prognosis or alter therapy. The correlation between cellularity and ADC was weaker in meningiomas (r = −.67) than in gliomas, but this also is of little concern. In meningiomas, information that improves our ability to predict which tumors are malignant or atypical would be important as this would affect therapeutic decisions. In a recent publication, Filippi et al (2) demonstrated that low ADC values (less than those in normal brain) were seen most frequently in malignant or atypical meningiomas. If this finding is confirmed in a larger series, it could prove to be an important advance in the diagnosis of meningiomas. Correlations between specific histologic features associated with malignancy and ADC should be sought.

Diffusion changes in metastatic disease also need further investigation. In this publication, the authors evaluated 21 patients with metastatic disease from four different primary tumors. There is, in reality, no reason to group all of these metastatic diseases in a single category. The histologic features of each tumor type is different and therefore one would expect that each type would have its own range of ADC values. Combining different tumor types only obscures potential differences between types. Further studies of cases in each primary tumor will be needed to determine if diffusion-weighted imaging has any role in the assessment of these lesions.

Initially, Kono et al's findings seem disheartening, but on further thought they offer possibilities. The fact that diffusion images, ADC maps, and even measurement of ADC values did not allow for differentiation between gliomas, meningiomas, and metastases is neither a surprise nor a major problem. Clinical information and routine “anatomic” MR imaging findings allow for correct diagnosis in the vast majority of intracranial neoplasms. Cases wherein these findings are discordant with the pathologic diagnosis represent rare diseases or unusual manifestations of common diseases. Only a truly magical test could prove definitive in these circumstances. Diffusion-weighted imaging is not magical, but it may yet provide valuable information about intracranial neoplasms that will improve our understanding of these diseases and aid in their diagnosis and treatment planning. We need to look at each disease process separately, gather more cases, and ask the right questions.

At the end of our protocol party, we had removed diffusion-weighted imaging from a few specific protocols (chronic seizures, multiple sclerosis follow-up). We all agreed that diffusion imaging should stay in our “rule out” protocols. After all, the art of clinical diagnosis remains just that, and diffusion-weighted imaging takes less than 1 minute to perform. Feeling that we had completed our task like responsible adults, we headed off to the tavern, the Nimrods, extending the holiday spirit one more day.

## References

1.  Tien RD, Felsberg GJ, Friedman H, Brown M, MacFall J. **MR imaging of high-grade cerebral gliomas: value of diffusion-weighted echo planar pulse sequences.** AJR Am J Roentgenol 1994;162:671-677
    
    [PubMed](http://www.ajnr.org/lookup/external-ref?access_num=8109520&link_type=MED&atom=%2Fajnr%2F22%2F6%2F1013.atom) 
    
    [Web of Science](http://www.ajnr.org/lookup/external-ref?access_num=A1994MY86000038&link_type=ISI) 

2.  Filippi CG, Edgar MA, Uluğ JC, Prowda JC, Heier LA, Zimmerman RD. **.** AJNR Am J Neuroradiol 2001;22:65-72
    
    [Abstract/FREE Full Text](http://www.ajnr.org/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6NDoiYWpuciI7czo1OiJyZXNpZCI7czo3OiIyMi8xLzY1IjtzOjQ6ImF0b20iO3M6MjA6Ii9ham5yLzIyLzYvMTAxMy5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

*   Copyright © American Society of Neuroradiology