| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters to the Editor |
Department of Radiology, Louisiana State University Medical Center, 1542 Tulane Avenue, Room 12, New Orleans, LA 70112-2822, e-mail: mpandi@lsuhsc.edu
Editor:
This letter is in response to the study conducted by Dr Swensen and colleagues and reported in the January 2000 issue of Radiology (1).
The findings of the multicenter study underscore the difficulty in compiling data from a multitude of scanners manufactured by different vendors. Moreover, there can be little hope of demonstrating the validity of the basic physical principle of photoelectric effect and the utility of dual-kilovolt scanning in the evaluation of solitary pulmonary nodules without reliable controls.
The success of applying dual-kilovolt computed tomography (CT) in the evaluation of solitary pulmonary nodules depends on a thorough and meticulous testing of the scanners. First, the scanner must be thoroughly assessed for zero errors and inhomogeneity caused by variations in beam hardening and scatter. These tests with a water phantom must be conducted with the same scanning parameters used for chest CT, with both kilovolt settings. Without this test, one cannot reliably use attenuation readings or even a change in readings. Furthermore, one must test the equipment periodically and after major maintenance work on the scanner. The authors failed to mention such tests in their article.
In addition, the authors failed to use soft-tissue controls in every patient. It is essential to measure the change in density of soft tissues at both kilovolt settings. A definite increase in density readings of calcified nodules may not be apparent, but when it is compared with the change in density readings of soft tissues, there is a relative increase in the density readings of the nodule. Again, these measurements must be made at both kilovolt settings. In the absence of soft-tissue controls, the reliance on absolute numbers, even a change in density readings, is likely to create errors.
Furthermore, if soft-tissue controls are used, the measurements must be performed in the same sector as the nodule so that the inherent inhomogeneity of scatter and beam hardening do not create errors in reading. Measurement in a corresponding sector would be critical, especially if the scanner has documented inhomogeneity.
Findings in the Japanese arm of the study performed with histologic correlation of the tissue calcium content are encouraging. The moderate correlation shown by the group could have been enhanced had the scanners been thoroughly checked for zero errors and inhomogeneity and had soft-tissue controls been used. Nevertheless, their results support the validity of the basic physical principle of the photoelectric effect.
REFERENCES
Department of Radiology, Mayo Clinic, 200 First St SW, Rochester, MN 55905, e-mail: mccollough.cynthia@mayo.edu
We agree with Dr Pandit-Bhalla that the successful application of any quantitative CT technique depends on both thorough and meticulous testing of the CT scanners, and we appreciate this opportunity to provide additional information regarding the rigorous quality assurance testing performed at our institution (Mayo Clinic, Rochester, Minn, and Mayo Clinic, Scottsdale, Ariz). Of the 240 nodules examined in this study (1), 210 were scanned at a Mayo Clinic site; thus, the findings from the two locations have the predominant influence on the reported results and conclusions. Our co-authors from the Shiga Health Insurance Hospital, Otsu, Japan (n = 25) and the Duke University Medical Center, Durham, NC (n = 5) reported that quality control at their institutions is performed in a similar manner.
The Department of Radiology at the Mayo Clinic has a long-standing history of both expertise in and commitment to high levels of quality assurance with all of their imaging devices, particularly with CT systems (2–10). Our medical physicists and quality-control technologists ensure that all of our systems not only meet but exceed applicable state and federal regulations, as well as professional recommendations (11–14). A complete quality assurance performance evaluation is performed with all of our CT systems at least biannually (our state regulations require only annual evaluations). In addition, we retest systems after any major equipment (eg, x-ray tube) or software changes.
In addition to the tests performed by our medical physics and quality-control staff, our in-house service engineers perform biweekly or monthly preventative maintenance tests (dependent on system model). Tests of image uniformity and the CT number of water are among those performed with all enabled peak kilovolt settings. Our service records indicate that calibration and inhomogeneity errors beyond the manufacturer acceptance limits are rarely observed (manufacturer limits: the mean CT numbers of water and center-to-edge difference must both be 0 HU ± 3). Moreover, even subtle shifts in the data (eg, data are within acceptance limits but have shifted away from the mean value) are corrected by our service engineers so that our scanners exhibit high levels of consistency. Thus, we are confident that calibration errors and inhomogeneity did not play a substantial role in the reported results.
Although not performed at the time of our study (1), scanning of a water phantom is now performed every morning prior to patient examinations. The daily water image is examined for the absence of artifacts, appropriate CT number calibration, and image noise. A service call is immediately initiated for any out-of-specification finding, and patient scanning is delayed if the artifacts or water calibration and noise values are of substantial concern. (The frequency of this test was increased from biweekly to daily because of state requirements, not because of any sense of equipment instability.)
Among 337 measurements recently acquired with the same equipment used in our study (1), only one data point had an unacceptable mean CT number of water (-6.3 HU, state of Minnesota limits are ±5 HU). The mean and SD values were -1.26 HU ± 1.22 (n = 337). Thus, the same equipment and calibration procedures as those used in our study show variations of only a few Hounsfield units in the mean CT number of water. These data confirm that potential drifts in the CT number calibration of our equipment are not a relevant issue. The practice of obtaining daily water scans, which we highly recommend for all CT-scanning sites, helps us to identify scanner problems on phantom images, where they are obvious, rather than on patient images, where they may be subtle or cause a need for a patient to be rescanned.
Despite these rigorous efforts to maintain accurate calibrations, daily variations can of course occur. Air calibrations are performed daily to further minimize CT number fluctuations. These frequent air calibrations are recommended by the manufacturer and are often automatically performed with the CT system software. Thus, even sites performing less frequent quality assurance testing will likely still be performing frequent air and water calibrations by virtue of the manufacturer recommendations or automated programs.
The typical level of CT number variations (of water) was discussed in the concluding paragraphs of our article (1), where we estimated that errors on the order of 0–6 HU can be expected. Thus, any quantitative CT technique would require CT number changes of approximately 20–30 HU for the reproducible detection of statistically significant changes. We agree that, in every patient, the use of some type of control, such as an appropriate anatomic region or calibration phantom, would be an appropriate method to further reduce potential inaccuracies in the CT numbers obtained at each of the kilovolt settings. This technique was not incorporated in our study. Bhalla et al report in their previous article (15) a mean decrease in CT numbers (
16–21 HU) between 140 and 80 kVp in regions of interest placed over soft tissue. They do not, however, describe the manner in which these data were used, if at all, to correct noted changes in CT numbers in nodules. Further, as x-ray attenuations from soft tissue and bone are produced predominately by different mechanisms (Compton and photoelectric interactions, respectively), a calcium-containing phantom, rather than patient soft tissue, may be a more appropriate control.
We would further like to point out that the level of calcium noted in the 15 histologically analyzed nodules in our study was in the range of 50–200 µg/g (1). By assuming a density of 1 g/mL, this calcium level is equivalent to a calcium content in the range of 0.05–0.20 mg/mL, which is well below the range that Bhalla et al (15) found to produce statistically distinguishable changes in CT numbers between the 80- and 140-kVp images. This finding is consistent with our reported results that the calcium content of indeterminate nodules (ie, not obviously calcified), as reflected by a change in mean CT number between 80- and 140-kVp images, is sufficiently small that it cannot be accurately distinguished from measurement noise.
In conclusion, we too are disappointed that the dual–kilovolt peak technique does not appear to be of substantial clinical value in the identification of benign versus malignant solitary pulmonary nodules. The fundamental physical principles of photoelectric and Compton interactions are in no way invalidated by our reported results. Rather, it seems that the confounding factors of polyenergetic x-ray spectra, the existence of calcified malignant and noncalcified benign nodules, and the statistical uncertainty inherent in the measurement of CT numbers mask the expected physical effects of increased CT numbers for calcified objects at lower kilovolt peaks.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
