100 days with scans of the same Catphan phantom on the same CT scanner

Abstract Quality control (QC) of CT scanners is important to evaluate image quality and radiation dose. Different QC phantoms for testing image quality parameters on CT are commercially available, and Catphan phantoms are widely used for this purpose. More data from measured image quality parameters on CT are necessary to assess test methods, tolerance levels, and test frequencies. The aim of this study was to evaluate the stability of essential image quality parameters for axial and helical scans on one CT scanner over time. A Catphan 600 phantom was scanned on a Philips Ingenuity CT scanner for 100 days over a period of 6 months. At each day of testing, one helical scan covering the entire phantom and four axial scans covering four different modules in the phantom were performed. All images were uploaded into Image Owl for automatic analysis of CT numbers, modular transfer function (MTF), low‐contrast resolution, noise, and uniformity. In general, the different image quality parameters for both scan techniques were stable over time compared to given tolerance levels. Average measured CT numbers differed between axial and helical scans, while MTF was almost identical for helical and axial scans. Axial scans had better low‐contrast resolution and less noise than helical scans. The uniformity was relatively similar for axial and helical scans. Most standard deviations of measured values were larger for helical scans compared to axial scans. Test results in this study were stable over time for both scan techniques, but further studies on different CT scanners are required to confirm that this also holds true for other systems.

evaluate the stability of essential image quality parameters for axial and helical scans on one CT scanner over time. A Catphan 600 phantom was scanned on a Philips Ingenuity CT scanner for 100 days over a period of 6 months. At each day of testing, one helical scan covering the entire phantom and four axial scans covering four different modules in the phantom were performed. All images were uploaded into Image Owl for automatic analysis of CT numbers, modular transfer function (MTF), low-contrast resolution, noise, and uniformity. In general, the different image quality parameters for both scan techniques were stable over time compared to given tolerance levels. Average measured CT numbers differed between axial and helical scans, while MTF was almost identical for helical and axial scans. Axial scans had better low-contrast resolution and less noise than helical scans. The uniformity was relatively similar for axial and helical scans. Most standard deviations of measured values were larger for helical scans compared to axial scans. Test results in this study were stable over time for both scan techniques, but further studies on different CT scanners are required to confirm that this also holds true for other systems. that only serious drifting or sudden errors are detected and not clinically irrelevant daily fluctuations. The recommended test frequency should aim to balance between being frequent enough to ensure adequate image quality, while infrequent enough to reduce time spent by the staff and avoid unnecessary interference with clinical work. 6,7 CT numbers, spatial resolution, low-contrast resolution, noise, and uniformity are important image quality parameters in QC of CT systems. 5,11 The CT number describes the x-ray attenuation characteristics of the material scanned relative to that of water. 11 Spatial resolution in CT refers to the ability to differentiate small high-contrast objects in an image. 5 In the scan plane, the spatial resolution can be described by the MTF. 11 Spatial resolution is important for example in visualizing iodine-enhanced vessels or small bone structures. 4 The low-contrast resolution refers to the ability to differentiate objects with slightly different density from the background. 5,11 Clinically, it is important for the ability to distinguish one soft tissue from another. 12 The noise is the standard deviation of a sample of CT numbers within a region of interest (ROI) in a homogenous part of a CT image. 11 To reveal beam hardening and cupping artifacts in CT images, uniformity is tested. Uniformity is a measure of the capability of the CT scanner to produce a uniform picture of a homogenous material. 5,7 Published guidelines are mainly based on image quality parameters evaluated from phantom images acquired in axial scanning mode. [8][9][10][11]13 However, in clinical practice, helical scans are mostly used based on advantages such as reduction of motion artifacts, scan time, and partial volume effects. 12 Automatic evaluation of QC images has the advantage of being observer independent and may also be time efficient. Storing of QC data in databases is a valuable tool to allow more in-depth analysis of failure rates of different test, changes that occur during the lifetime of the equipment and also to compare CT scanners. New developments in CT scanner technology, including important features that affect image quality, are rapidly emerging. It is therefore essential to update methods for QC to make sure the most clinically relevant parameters are being tested.
Discussions on optimal testing methods, tolerance levels, and test frequencies are ongoing in the scientific environment. [2][3][4][5][6][7] More data from frequent measurements over time provide valuable insight into error rates and day to day fluctuations for the different tests.
Together with exploring different test methods, this will help establishing better QC routines for CT scanners. Information on day to day fluctuations of measured image quality parameters is available in the literature, but to our knowledge no study has been performed with Catphan which is widely used by medical physicists. 6,7 Studies using both axial and helical scan technique for QC are also available, but to our knowledge no systematic comparison of results from the different techniques has been carried out. 4,6 Data from repeated helical and axial scans on the same CT and Catphan are therefore interesting to (a) compare helical and axial scan techniques and (b) to evaluate fluctuations in measured parameters from a Catphan.
The aim of this study was to evaluate the stability of essential image quality parameters for axial and helical scans on one CT scanner over time.

| METHODS
The same Catphan 600 phantom was scanned on a Philips Ingenuity CT (Philips Medical Systems, Best, Netherlands) for several days over a period of 6 months, resulting in a total of 100 scans. The CT scanner had been in clinical use for about one and a half years when this study was initiated. The workload on the CT scanner before and during this study was about 20 patients per day, 4-5 days a week. Air calibration was performed on the same day before initiating the scans. The phantom was accurately positioned by matching the markings on the outside of the phantom with the coronal, axial, and sagittal lasers on the CT scanner. The accuracy of the lasers was established before initiation and controlled during the course of the study. No changes to the lasers were made during this study.
The same specialized CT radiographer performed the majority of the scans, while the same medical physicist performed the scan in case of absence. To ensure that identical scan parameters were used each time, scan protocols were saved on the CT scanner. One helical scan and four axial scans were performed each time. The helical scan covered the entire Catphan. Center positions for the axial scans were at positions 0, À80, À110, and À160 mm along the z-axis, corresponding respectively to CTP404 sensitometry module (CT numbers), CTP528 high-resolution module (MTF), CTP515 low-contrast resolution module, and CTP486 uniformity module (noise and uniformity).
Tube current modulation and iterative reconstruction were not applied for any of the scans. The scan protocol was not based on a clinical exam protocol. The helical scan protocol used in this study was developed for standardized acceptance and annual testing, and is used by more than 30 different radiological departments in Norway. The axial scan protocol was designed to be as similar as possible to the helical scan protocol. Scan parameters used for helical and axial scans are presented in Table 1.
The images were uploaded into Image Owl (Image Owl, Inc., Greenwich NY, USA), where analyses of CT numbers, MTF, lowcontrast resolution, noise, and uniformity were performed automatically. Image Owl was also used to group the results from the different tests and dates in to trends. These trend data were then transferred to excel for further analysis. Statistical analysis was per- Levene's test was performed to determine if there was a statistically significant difference in variance between helical and axial scans (P < 0.05).

2.B | MTF
The high-resolution module (CTP528) contains a lower bead point source that was used for evaluation of MTF and the calculation of critical frequency values [cycles/cm] for 50% and 10% of MTF.

2.C | Low-contrast resolution
The low-contrast module (CTP515) contains supra-slice targets, which are cylindrical objects with diameters of 2-15 mm and having contrast levels of 0.3%, 0.5%, and 1.0%. It also has sub-slice targets, with z-axis dimensions of 3, 5, and 7 mm at 1.0% contrast, centered in the z-dimension of the module. In this study, the supra-slice targets were used for the low-contrast evaluation.

2.D | Noise
The image uniformity module (CTP486) is cast from a uniform material with typical CT number 5-18 HU. The noise was measured in Image Owl as the standard deviation of CT numbers in a central ROI with a diameter 40% of the diameter of the uniformity module.

3.A | CT numbers
Average values, maximum deviation from average values, absolute difference between minimum and maximum values (range), and standard deviations (r) of measured CT numbers for each material are presented in Table 2 for both axial and helical scans.
The axial scans had higher average CT numbers than the helical scans for all materials except air. The difference between axial and helical scans in average CT numbers also increased with increasing density for materials denser than air. The standard deviations of the measured CT numbers were larger for the helical scans than for axial scans for all materials except air. There was a statistically significant difference in variances between axial and helical scan technique for all materials except air. The absolute value of the slopes for the different materials and scan techniques were all <0.017. Figure 2 displays how the measured CT numbers for each of the different materials and scan techniques varies with time.

3.B | MTF
Average values, maximum deviation from average values, range, and standard deviations (r) of critical frequency values for 50% and 10% of MTF, are presented in Table 3 for axial and helical scans.
The average critical frequency values for 50% and 10% of MTF were almost identical for axial and helical scans. The maximum deviation, range, and the standard deviation of critical frequency values were generally smaller for the axial scans than the helical scans, T A B L E 1 Scan parameters used for helical and axial scans of Catphan 600 phantom.

3.C | Low-contrast resolution
Average values, range, and standard deviations (r) of estimated diameter [mm] of smallest detectable target for each of the three contrast values (1%, 0.5%, and 0.3%) are displayed in Table 4 for axial and helical scans.
The average low-contrast resolution was better (smaller detectable targets) for axial scans compared to helical scans for 0.5% and 0.3% contrast. For 1% contrast, the smallest possible target at 2 mm was detected with both scan techniques. The range and standard deviation were larger for helical scans than for axial scans where the measured values did not vary at all over time.

3.D | Noise
Average values, maximum deviation from average values, range, and standard deviations (r) of measured noise values are presented in Table 5 for axial and helical scans.

3.E | Uniformity
Average values, maximum deviation from average values, highest measured value, range, and standard deviation (r) of uniformity measurements are presented in Table 6 for both axial and helical scans.

4.B | MTF
MTF is most dependent on detector element size, reconstruction matrix, DFOV, scanner geometry, focus size, and reconstruction algorithm. 4 IPEM suggests a remedial level for yearly constancy control of baseline AE20%, but states no suspension level. 8 According to IEC, a tolerance level of baseline AE0.5 lp/cm or baseline AE15%, whichever is greater, is given for measurement of the 50% point and 10% point of  10 The measured MTF in this study was very stable over time for both axial and helical scans, where none of the maximum deviations from average values were above 5%. Significant shift in MTF over time is also not expected, even though high resolution scans might show some changes due to focal spot degradation in an aging tube. Software updates like modification of the reconstruction algorithm or hardware changes that affect detector or focus alignment can alter the MTF. 13 Roa et al evaluated QC data from different CT scanners retrospectively, and also concluded that the spatial resolution did not change much over time. 4 Calculated slopes showed no apparent trend over time and average values were almost identical between axial and helical scans. Spatial resolution should be comparable between axial and helical scans if the same acquisition parameters are used. 11

4.C | Low-contrast resolution
Low-contrast resolution is influenced by signal to noise ratio (SNR), the spatial resolution and the reconstruction algorithm. 11 IPEM do not include low-contrast resolution in the recommended test regime. 8 According to IEC, low-contrast resolution is not a necessary part of acceptance or constancy QC, as if measured noise and CT numbers meet specifications, the low-contrast resolution is also deemed to meet specifications. 9,10 The results from our study showed no variations at all in results for axial scans, while helical scans generally had small variations and slightly worse low-contrast resolution compared to axial scans. This is consistent with less measured noise in axial scans than helical scans.

4.D | Noise
Noise is a good first-line measurement, as several different parameters influence noise (like kV, mAs, filtration, slice width, reconstruction algorithm, image filter, and focal spot size), which means that measured deviations can arise from a number of different sources. 13 IPEM suggest noise measurements to be performed daily to weekly and annually with a remedial level of baseline AE10% and a suspension level of baseline AE25%. 8 The tolerance level stated by IEC is baseline AE10% or 0.2 HU, whichever is larger, and a test frequency of at least monthly. 10 The measured noise in this study was very stable over time for both axial and helical scans, none of the maximum deviations from average values were above 5%. The stability of noise found in this study is also consistent with a 2 yr automatic monitoring of QC parameters by Nowik et al Over this period with daily scans, noise levels were only once measured outside their tolerance level of 5% (deviation just above 10%) caused by a barely visible ring artifact. 7

4.E | Uniformity
The shaped beam filter, x-ray tube output, and centering of the object in the beam can influence the uniformity of an image. 7  deviation in water. The CT number of water, although not with the highest failure rate (estimated failure rate 0.85 failures/1000 scan days), is important due to increased use of quantitative imaging. 6 Dose measurements and testing of automatic dose modulation should be part of the QC regime of a CT scanner, however, this study only addresses image quality testing, hence, testing of dose and automatic dose modulation is beyond the scope of this study.

| CONCLUSION
The results from this study showed stability over time for all image quality tests for helical and axial scan techniques. However, these results are based on just one CT scanner, and even though other studies have also shown similar tendencies, further studies on different CT scanners are necessary. Additional studies are planned. It is also essential with continued research to develop the most time-efficient and clinically relevant QC methods and imaging phantoms.

CONF LICT OF I NTEREST
Oslo University Hospital has a research collaboration with Image Owl and The Phantom Laboratory.