We read with interest the article on subtraction angiography of the intra- and extracranial vessels by Jayakrishnan et al (1) in the March 2003 issue of the AJNR. From the authors’ description, we surmise that their technique is a variant of a technique we have routinely used in our hospital for more than 4 years.
We would like to make the following comments. As in all subtraction techniques, two scans are made: one precontrast and the other postcontrast. We understand, however, that the first scan is used to identify the high attenuation structures (ie, bones and calcifications) and that this information is used to mask these structures in the postcontrast scan. If this assumption is correct, the term “subtraction” for this procedure is understandable—actually we used the same phrase in our first communication on this subject (2)—but it is incorrect. In subtraction, all the pixels in the volume of interest are involved, which leads to an overall deterioration of the image quality because of the increase of image noise, whereas in masking only the CT values of the high attenuation pixels are affected. This difference is especially important when the precontrast scan is made with a low dose. In an article on this subject (3), we explained this last method (matched mask bone elimination, or MMBE) and clearly demonstrated the advantages of masking over subtraction.
The use of a vacuum-type head holder to minimize movement is an interesting addition to this technique. The authors state that the use of this head holder facilitates image registration. From the information given in the rest of this article, we assume, however, that no registration is performed at all. In our experience, even the slightest movement of the patient in between the scans (in the order of 0.05 mm) may lead to a serious degradation in image quality. If (as we assume) the authors do not use any registration, it would be interesting to investigate whether addition of such a registration step would produce an improvement in the quality of the processed images.
The use of image registration is completely feasible in a routine clinical setting. The authors give a mean postprocessing time for their procedure of slightly more than 8 minutes, by using software of the Omipro workstation of the CT-Twin CT-scanner. For comparison, the processing time of the MMBE software initially was in the order of 1 hour for one examination (3); because of improvements of the software and better performance of the hardware, this time has now been reduced to less that 10 minutes.
Reply:
We would like to thank Drs. Venema and den Heeten for their comments on our article (1).
We used the term “subtraction” in a manner consistent with the terminology used in the Omnipro workstation to describe the process of subtracting a 3D mask from the 3D maximum intensity projection image. This does not mean a pixel-by-pixel subtraction, which will of course result in increased noise.
We did not use any image registration protocol in this study, because we often scan the head and neck and the degree of movement generated are probably more than what can be handled by such registration methods. It is for this reason we highlighted the use of the vacuum bag, which is a very good method of immobilization and is well tolerated by the patients. We agree with the Drs. Venema and den Heeten in that even the slightest movement between the scans can seriously degrade the image quality. To counter this, a further process of 3D mask expansion was sometimes employed, usually a one-pixel expansion was sufficient to reduce the artifacts to an acceptable level—a process similar to that used by Venema et al. (2).
References
- Copyright © American Society of Neuroradiology