Introduction NEURO. Birgit Ertl-Wagner Lara Eftimov Jeffrey Blume Roland Bruening Christoph Becker Jean Cormack Hartmut Brueckmann Maximilian Reiser

Transcription

1 Eur Radiol (2008) 18: DOI /s NEURO Birgit Ertl-Wagner Lara Eftimov Jeffrey Blume Roland Bruening Christoph Becker Jean Cormack Hartmut Brueckmann Maximilian Reiser Cranial CT with 64-, 16-, 4- and single-slice CT systems comparison of image quality and posterior fossa artifacts in routine brain imaging with standard protocols Received: 22 April 2007 Revised: 21 January 2008 Accepted: 25 January 2008 Published online: 4 April 2008 # European Society of Radiology 2008 B. Ertl-Wagner (*). L. Eftimov. C. Becker. M. Reiser Institute of Clinical Radiology, University of Munich Campus Grosshadern, Marchioninistr. 15, Munich, Germany B.Ertl-Wagner@t-online.de Tel.: Fax: J. Blume. J. Cormack Center for Statistical Sciences, Brown University, Providence, RI, USA R. Bruening. H. Brueckmann Department of Neuroradiology, University of Munich Campus Grosshadern, Munich, Germany Abstract Posterior fossa artifacts constitute a characteristic limitation of cranial CT. To identify practical benefits and drawbacks of newer CT systems with reduced collimation in routine cranial imaging, we aimed to investigate image quality, posterior fossa artifacts and parenchymal delineation in non-enhanced CT (NECT) with 1-, 4-, 16- and 64-slice scanners using standard scan protocols. We prospectively enrolled 25 consecutive patients undergoing NECT on a 64- slice CT. Three groups with 25 patients having undergone NECT on 1-, 4- and 16-slice CT machines were matched regarding age and sex. Standard routine CT parameters were used on each CT system with helical acquisition in the posterior fossa; the parameters varied regarding collimation and radiation dose. Three blinded readers independently assessed the cases regarding image quality, infraand supratentorial artifacts and delineation of brain parenchymal structures on a five-point ordinal scale. Reading orders were randomized. A proportional odds model that accounted for the correlated nature of the data was fit using generalized estimating equations. Posterior fossa artifacts were significantly reduced, and the delineation of infratentorial brain structures was significantly improved with the thinner collimation used for the newer CT systems (p<0.001). No significant differences were observed for midbrain structures (p>0.5). The thinner collimation available on modern CT systems leads to reduced posterior fossa artifacts and to a better delineation of brain parenchyma in the posterior fossa. Keywords Posterior fossa. Image quality. Artifacts Introduction Since the introduction of multi-slice technology in the year 1998 by several major CT vendors, computed tomography (CT) has undergone a rapid further development [1 3]. Currently, modern generation CT systems are able to simultaneously acquire 64 parallel slices, thus allowing a previously unparalleled spatial and temporal resolution [4 8]. With the advent of multi-slice technology with higher numbers of simultaneously acquired slices, several prerequisites of cranial CT changed [9, 10]. Depending on the manufacturer, it became impossible to angulate the gantry for some CT machines with high numbers of simultaneously acquired slices. However, this problem can be at least partially counteracted by appropriate patient positioning within the head rest. In many clinical settings, it

2 1721 became standard practice to acquire a spiral CT volume acquisition of the entire brain with subsequent multiplanar reformations in an axial and often also in a coronal and sagittal-orientation, when CT systems with high numbers of simultaneously acquired slices are used. Another option is to obtain sequential images with a thin collimation. Posterior fossa artifacts have always been viewed as a major limitation of cranial CT; these often hinder the delineation of hindbrain structures such as the brainstem and/or the cerebellum. CT studies of hindbrain structures were frequently considered uninterpretable due to pronounced artifacts [11, 12]. Small infarcts or metastases of hindbrain structures therefore tended to be a diagnostic domain of MR imaging. The introduction of spiral as compared to sequential CT first led to an initial decrease of posterior fossa artifacts (13); cranial CT images were therefore commonly acquired in a spiral mode in the posterior fossa and in a sequential mode for the supratentorial structures (9, 10). The gantry was commonly angulated parallel to the skull base in order to avoid radiation exposure to the eye lens with no demonstrated effect on posterior fossa artifacts (11, 12). In the year 2001, Jones et al. demonstrated that a CT of the brain could be acquired 1.8 times faster with 4-slice as compared to single-slice technology (14). Posterior fossa artifacts were considered less pronounced with 4-slice technology as compared to single-slice technology in 94% of cases, and the overall reader preferences lay with the 4-slice as compared to the single-slice technology in 89% (14). From a theoretical standpoint, the introduction of newer CT systems with novel focus technology and thinner collimations may have the potential to further reduce posterior fossa artifacts due to their improved z-sampling (4, 5). However, the sheer increase in the number of slices may also have a negative influence on posterior fossa artifacts, since cone beam artifacts increase with an increased cone angle. This especially holds true in regions with horizontally oriented high-contrast structures such as the posterior fossa. The aim of our multi-reader, matched-control study was to identify practical benefits and drawbacks of various CT scanner types by comparing parameters of posterior fossa artifacts, image quality and delineation of cerebral structures, between single-slice scanners and scanners with 4, 16 and 64 simultaneously acquired slices, respectively. Materials and methods Patient selection Expedited Institutional Review Board (IRB) approval was obtained. We enrolled 25 consecutive patients referred for a nonenhanced cranial CT into the study. All 25 patients underwent cranial CT on a 64-slice system. Inclusion criteria were an age over 18 and an indication and referral for a nonenhanced CT of the brain. Exclusion criteria consisted of an inability to fully cooperate with the CT examination due to an impaired patient compliance. Included patients were subsequently categorized according to sex and age. Patients examined on the 64-slice CT machine were then matched to patients examined on 4- and 16-slice CT systems and on a single-slice CT system, respectively. Matching was performed retrospectively. Consecutive patients having undergone cranial CT with the various CT systems were screened retrospectively and included into the study when age and sex criteria matched. Thus, 100 patients were included into the study with 25 patients for each type of CT systems. After completion of patient accrual a consensus panel of two radiologists who were not involved in the study as readers reviewed all cases. Cases with significant motion artifacts were excluded from the study, and the next consecutive matching patient was included instead. Imaging protocols Patient examinations were performed on 64-slice CT (Somatom Sensation 64, Siemens Medical Solutions), on 16-slice CT (Somatom Sensation 16, Siemens Medical Solutions), on 4-slice CT (Somatom Sensation 4, Siemens Medical Solutions), and on single-slice spiral CT (Somatom Plus 4, Siemens Medical Solutions), respectively. All examinations were performed with routine CT protocols. Single-slice CT was performed in a spiral mode for the infratentorial structures and in a sequential mode for the supratentorial structures. All other examinations were performed in a spiral mode. Table 1 summarizes the various protocol parameters including kv and mas settings, collimation, pitch and reformations. All images were viewed in a standardized axial orientation. Image evaluation Images were assessed by three blinded readers. Readers were masked regarding any patient identifying information, any patient demographic parameters, the type of CT system used and the acquisition parameters. Readings were performed independently, and reading orders were randomized for each reader. Two readers were board-certified attending neuroradiologists holding a certificate of added qualification in neuroradiology, and one reader was a board-certified attending radiologist with a special expertise in computed tomography. All data sets were evaluated on a standard PACS workstation (Siemens Magic View, Siemens Medical Solutions). Window and level settings were standardized and kept constant for all examinations.

3 1722 Table 1 Imaging protocols for the various CT systems used in the study *Values for the posterior fossa and the skull base; for the supratentorial brain, 8-mm slice width and collimation, and 360 mas were selected S1 S4 S16 S64 KV mas 180* Slice width 4.0 * 5.0 mm 5.0 mm 5.0 mm Collimation 2.0 * 2.5 mm 1.5 mm 0.6 mm Rotation time 1.5 s 0.75 s 1.0 s 1.0 s Feed/rotation 2.0 mm 6.5 mm 13.2 mm 16.3 mm Pitch Reconstruction increment 2.0 mm 5.0 mm 5.0 mm 5.0 mm Field of view 200 mm 200 mm 200 mm 200 mm Reconstruction kernel AH30 H20s smooth H20s smooth H31s smooth The readers evaluated the images with a standardized questionnaire. Overall image quality was rated on a 5-point ordinal scale with five corresponding to an optimal image quality and one corresponding to a poor image quality hindering image interpretation. The readers also rated the presence of artifacts in the posterior fossa as well as the presence of supratentorial artifacts on a 5-point ordinal scale with five corresponding to a total absence of artifacts and one corresponding to major artifacts hindering image interpretation. In addition, readers were asked to rate the delineation of the following cerebral structures on a 5-point ordinal scale with five corresponding to perfect delineation of the structure in its entirety and one corresponding to a structure that cannot be delineated at all: medulla oblongata, pons, cerebellum, midbrain and temporal lobe. Statistical analysis Comparisons of image quality, presence of artifacts and the delineation of the various cerebral structures across the four types of CT systems, as graded by each of three blinded readers on a 5-point scale, were carried out in a two-stage process. In the first stage, for exploratory purposes, we assessed an average grade (despite the ordinal nature of the data) for each patient for each CT system, averaging across the three readers. A Kruskal-Wallis non-parametric rank F-test was then used to test the null hypothesis that the mean rank was the same across the four scanner types. Advantages of the non-parametric nature of the Kruskal-Wallis test are that no distributional assumptions are required for the data; furthermore, the slightly conservative nature of the test allows us to narrow the number of anatomic structures and image parameters down to those likely exhibiting true differences among the different types of CT machines. This first stage was used to determine which parameters exhibited significant differences among the four types of CT systems. In the second stage, significant differences above were analyzed more closely through means of a proportional odds multinomial regression model for ordinal data. All observations for each case, corresponding to the three readers, were used in fitting the model, with the use of generalized estimating equations (GEE) to adjust all standard errors for the correlation existing between grades given by different readers for the same case. The key advantage of the proportional odds model is that it allows one to take advantage of the ordinal nature of the data; wherein, we are modeling cumulative logits of the probability of a response, while adjusting for explanatory variables. The assumption made is that the influence of explanatory variables, in this case the type of CT system or the protocol applied respectively, is independent of the cutpoint for the cumulative logit. For example, in comparing scanner type A versus scanner type B, the odds ratio for the probability of a grade of 5 versus the probability of a grade of less than 5 is the same as the odds ratio for the probability of a grade of greater than or equal to 3 versus the probability of a grade of less than 3. The null hypothesis of equality of parameter estimates from the proportional odds model among the four types of scanners was assessed through use of a score test with 3 of freedom, resulting from the GEE. Also, odds ratios and 95% confidence intervals for each pair-wise comparison of scanner types and scan protocols were then calculated and reported. Results A total of 64 men (16 per group) and 36 women (9 per group) were included in the study. There were no significant differences regarding age and sex distribution between the groups. Table 2 summarizes the demographic data of the patients examined with the various scanner types. Table 3 demonstrates the results of the Kruskal-Wallis non-parametric rank F-test representing the first stage of analysis, while Tables 4 and 5 summarize the results of the proportional odds multinomial regression model for ordinal data providing the significance levels and the

4 1723 Table 2 Mean age and sex of patients examined with the various CT systems Total no. of patients Men Women Mean age in years 53.7± ±10.9 Mean age S ± ±11.7 Mean age S ± ±11.2 Mean age S ± ±11.4 Mean age S ± ±11.2 odds ratios (OR) with 95% confidence intervals for the comparisons between the respective scanner types for each pair-wise comparison. Image quality overall was significantly improved from the single- to the 4-slice machine, as well as from the 16- to the 64-slice systems, while there were no significant differences between the 4- and the 16-slice machines. Artifacts in the posterior fossa were significantly reduced from 4- to 16-slices as well as from 16- to 64- slices, respectively, while there were no significant differences between the 4-slice and the single-slice examination (Figs. 1, 2, 3, 4). Supratentorial artifacts were significantly reduced on the 4-slice system compared with the single-slice systems and on the 64-slice system compare with 16-slice system. There were no significant differences between the 4- and the 16- slice examinations. The medulla oblongata, the pons and the cerebellum could be significantly better delineated on the 64- as compared to the 16-slice examinations, and with 16- as compared to the 4-slice examinations, while there were no significant differences between the 4- and the single-slice results. No significant differences in parenchymal delineation were observed when evaluating the midbrain structures. Table 3 Results of the Kruskal-Wallis non-parametric rank F-test across all parameters representing the first stage of analysis Kruskal-Wallis score P value Image quality 14.1 (3) Artifacts posterior fossa 52.1 (3) < Artifacts supratentorial 18.9 (3) Medulla oblongata 42.4 (3) < Pons 28.4 (3) < Cerebellum 28.9 (3) < Midbrain 2.3 (3) Temporal lobe 31.2 (3) < Basal ganglia 28.1 (3) < Cortex 31.9 (3) < The temporal lobes were significantly better delineated with 64- as compared to 16-slice examinations (p<0.05, OR 2.29), while there were no significant differences between 16- and the 4-slice studies (p>0.05, OR 1.87) and between 4- and single-slice examinations (p>0.05, OR 1.64). Discussion Our study demonstrates that increased image quality can be obtained with the use of thinner collimation on more modern multi-detector CT systems. The presence of artifacts in the posterior fossa was significantly reduced with high odds ratios and low p-values when comparing multi-detector CT of the brain acquired with a thinner collimation on CT machines with multiple simultaneously acquired slices with those acquired using thicker collimation and fewer simultaneously acquired slices. This reduction of artifacts in the posterior fossa can probably be in large part explained by the increasingly reduced slice collimation. In addition, novel focus technologies and data reformation from helical data sets may play a supplementary role [1, 4, 6, 15]. In what seems at a first glance a somewhat contradictory result to the literature, we found no significant decrease in posterior fossa artifacts when comparing single-slice CT to 4-slice CT [14]. However, Jones et al. compared 5-mm single-slice CT examinations with mm multi-slice studies. They therefore used a thinner collimation in the 4- slice CT compared to the single-slice CT, while we used a slightly thinner collimation in the posterior fossa (2 mm) in our single-slice CT as compared to our 4-slice CT (2.5 mm). This reduced collimation in the single-slice CT in our study explains the equal results when comparing posterior fossa artifacts to the 4-slice CT. As expected from the literature, supratentorial artifacts were reduced with 4-slice CT as compared to single-slice spiral CT [1, 2, 8, 14, 15] with a further reduction when going from 16-slice to 64-slice CT in our study. In a recent study, van Straten et al. have demonstrated that thinly collimated spiral CT with image combining leads to a better visualization of brain tissue near the skull when compared to broadly collimated sequential CT [15]. This is in accordance with our data, which overall also demonstrated reduced artifacts in regions in close proximity to the skull with reduced collimation. In accordance with our data on reduced artifacts in the posterior fossa, we found a significantly better delineation of infratentorial anatomical structures such as the medulla oblongata, the pons and the cerebellum when the slice number was increased from 4 to 16, and from 16 to 64 slices.this increased delineation of hindbrain structures can probably be in large part attributed to the abovementioned reduction in artifacts and therefore to the increasingly reduced slice collimation with a potential

5 1724 Table 4 Results for the pairwise comparisons of the various CT systems using a proportional odds multinomial regression model >>p<0.001>p<0.05 =p>0.05 Parameter Overall image quality S 64 > S16 = S4 > S1 Reduction of supratentorial artifacts S 64 > S16 = S4 >> S1 Reduction of posterior fossa artifacts S 64 > S16 >> S4 = S1 Cerebellum S 64 > S16 > S4 = S1 Pons S 64 > S16 >> S4 = S1 Medulla oblongata S 64 > S16 >> S4 = S1 Temporal lobes S 64 > S16 = S4 = S1 Midbrain S 64 = S16 = S4 = S1 supplementary role of novel focus technologies and data reformation [1, 4, 6, 11 13, 15].Odds ratios were especially high for the medulla oblongata, a region notoriously difficult to delineate in single-slice studies. In comparison to the infratentorial anatomical structures, imaging of the midbrain appears to be largely unaffected by increasing slice numbers with no significant differences in the delineation between the different groups. The delineation of the temporal lobes was also largely unaffected when comparing single-slice to 4-slice CT, and 4- to 16-slice CT. However, when comparing 64- to 16-slice CT, a significantly improved delineation was found. This effect is most likely also due to a reduced presence of artifacts in this region with a close proximity to the bony structures of the middle cerebral fossa and is probably again due to a decreased collimation. Even though MR imaging plays an important role in the evaluation of posterior fossa pathologies, we are convinced that CT still is and probably will remain for some time the mainstay in the emergency evaluation of neurological emergencies, e.g., in stroke patients. This is mostly due to the limited availability of MR imaging especially at night and during weekends. Moreover, patient access is limited in MR imaging, which often restricts its use in critically ill Table 5 Odds ratios with 95% upper and lower confidence bounds and p-values for each pair-wise comparison between CT systems OR: odds ratio, CI: confidence interval, LB: lower bound, UB: upper bound, p.f.: posterior fossa, supra: supratentorial Parameter Comparison Reference OR 95% CI LB 95% CI UB P value Image quality Image quality Image quality Artifacts p.f Artifacts p.f < Artifacts p.f Artifacts supra Artifacts supra Artifacts supra Medulla obl Medulla obl Medulla obl Pons Pons Pons Cerebellum Cerebellum Cerebellum Midbrain Midbrain Midbrain Temporal lobe Temporal lobe Temporal lobe

6 1725 Fig. 1 Infratentorial section of an unenhanced cranial CT scan acquired with a single-slice CT in a 68-year-old male patient stroke patients. Even though MR imaging is of undisputed value and in many instances is superior to CT, we believe that non-enhanced cranial CT will remain the mainstay in the emergency workup of acutely ill neurological for quite some time. Contrast-enhanced CT methods such as CT perfusion and/or CT angiography may further augment the diagnostic value of non-enhanced cranial CT. Further studies would need to show the impact of more modern scanner types in these methods. Our study has several limitations that need to be taken into account when interpreting the results. First, the four groups of patients examined with the various CT systems were not identical, as we considered it unethical to expose Fig. 3 Infratentorial section of an unenhanced cranial CT scan acquired with a 16-slice MS-CT in a 60-year-old female patient the patients to a higher than clinically necessary dose of radiation. In order to overcome this limitation as far as possible, we, however, employed a matched-control design to the study with a matching regarding age and sex across the groups. Second, we purposefully did not focus on various disease categories, but rather conducted a study of image quality, presence of artifacts and delineation of anatomical structures. As a myriad of disease categories exist that are relevant for cranial CT, a comprehensive study of all disease categories did not seem feasible. Further, focused studies on various disease categories, such as cerebellar or pontine infarcts or infratentorial metastases, with a dif- Fig. 2 Infratentorial section of an unenhanced cranial CT scan acquired with a 4-slice MS-CT in a 69-year-old male patient Fig. 4 Infratentorial section of an unenhanced cranial CT scan acquired with a 64-slice MS-CT in a 60-year-old female patient

7 1726 ferent study design will need to demonstrate the diagnostic potential of non-enhanced cranial CT with higher numbers of simultaneously acquired slices. Third, the protocols we used for the different CT machines are not identical. This is, however, at least in part a limitation inherent to the technical nature of the various CT systems. We used the CT protocols that were routine standard in our institution in order to attain a degree of comparability relevant to clinical practice. Fourth, we only compared CT machines from a single vendor. This was a limitation inherent to the situation at our institution at the time of the study, with four scanner types with different slice numbers in concomitant use. We believe, however, that, while certain aspects of the results may vary from vendor to vendor due to different focus and detector technologies, the general outcome with reduced artifacts and an improved delineation for infratentorial structures will hold true for the various vendors of multislice CT. In conclusion, our multi-reader, matched-control study demonstrated an overall improved image quality, reduced infra- and supratentorial artifacts and an improved delineation of infratentorial parenchymal structures with newer CT machines when standard protocols with an increasingly reduced slice collimation are used. References 1. Hu H, He HD, Foley WD, Fox SH (2000) Four multidetector-row helical CT: image quality and volume coverage speed. Radiology 215: Klingenbeck-Regn K, Schaller S, Flohr T, Ohnesorge B, Kopp AF, Baum U (1999) Subsecond multi-slice computed tomography: basics and applications. Eur J Radiol 31: McCollough CH, Zink FE (1999) Performance evaluation of a multi-slice CT system. Med Phys 26: Flohr TG, Stierstorfer K, Ulzheimer S, Bruder H, Primak AN, McCollough CH (2005) Image reconstruction and image quality evaluation for a 64-slice CT scanner with z-flying focal spot. Med Phys 32: Flohr T, Stierstorfer K, Raupach R, Ulzheimer S, Bruder H (2004) Performance evaluation of a 64-slice CT system with z-flying focal spot. Rofo 176: Flohr TG, Schaller S, Stierstorfer K, Bruder H, Ohnesorge BM, Schoepf UJ (2005) Multi-detector row CT systems and image-reconstruction techniques. Radiology 235: Napoli A, Fleischmann D, Chan FP et al (2004) Computed tomography angiography: state-of-the-art imaging using multidetector-row technology. J Comput Assist Tomogr 28(Suppl 1):S32 S45 8. Fleischmann D, Rubin GD, Paik DS et al (2000) Stair-step artifacts with single versus multiple detector-row helical CT. Radiology 216: Bahner ML, Reith W, Zuna I, Engenhart- Cabillic R, van Kaick G (1998) Spiral CT vs incremental CT: is spiral CT superior in imaging of the brain? Eur Radiol 8: Cody DD, Stevens DM, Ginsberg LE (2005) Multi-detector row CT artifacts that mimic disease. Radiology 236: Yeoman LJ, Howarth L, Britten A, Cotterill A, Adam EJ (1992) Gantry angulation in brain CT: dosage implications, effect on posterior fossa artifacts, and current international practice. Radiology 184: Rozeik C, Kotterer O, Preiss J, Schutz M, Dingler W, Deininger HK (1991) Cranial CT artifacts and gantry angulation. J Comput Assist Tomogr 15: Dorenbeck U, Finkenzeller T, Hill K, Feuerbach S, Link J (2000) Volumeartifact reduction technique by spiral CT in the anterior, middle and posterior cranial fossae. Comparison with conventional cranial CT. Rofo 172: Jones TR, Kaplan RT, Lane B, Atlas SW, Rubin GD (2001) Single- versus multi-detector row CT of the brain: quality assessment. Radiology 219: van Straten M, Venema HW, Majoie CB, Freling NJ, Grimbergen CA, den Heeten GJ (2007) Image quality of multisection CT of the brain: thickly collimated sequential scanning versus thinly collimated spiral scanning with image combining. Am J Neuroradiol 28: