Abstract
Background
Although CT urography (CTU) is widely used for the evaluation of the entire urinary tract, the most important drawback is the radiation exposure.
Purpose
To evaluate the effect of a noise reduction filter (NRF) using a phantom and to quantitatively and qualitatively compare excretory phase (EP) images using a low noise index (NI) with those using a high NI and postprocessing NRF (pNRF).
Material and Methods
Each NI value was defined for a slice thickness of 5 mm, and reconstructed images with a slice thickness of 1.25 mm were assessed. Sixty patients who were at high risk of developing bladder tumors (BT) were divided into two groups according to whether their EP images were obtained using an NI of 9.88 (29 patients; group A) or an NI of 20 and pNRF (31 patients; group B). The CT dose index volume (CTDIvol) and the contrast-to-noise ratio (CNR) of the bladder with respect to the anterior pelvic fat were compared in both groups. Qualitative assessment of the urinary bladder for image noise, sharpness, streak artifacts, homogeneity, and the conspicuity of polypoid or sessile-shaped BTs with a short-axis diameter greater than 10 mm was performed using a 3-point scale.
Results
The phantom study showed noise reduction of approximately 40% and 76% dose reduction between group A and group B. CTDIvol demonstrated a 73% reduction in group B (4.6 ± 1.1 mGy) compared with group A (16.9 ± 3.4 mGy). The CNR value was not significantly different (P = 0.60) between group A (16.1 ± 5.1) and group B (16.6 ± 7.6). Although group A was superior (P < 0.01) to group B with regard to image noise, other qualitative analyses did not show significant differences.
Conclusion
EP images using a high NI and pNRF were quantitatively and qualitatively comparable to those using a low NI, except with regard to image noise.
CT urography (CTU) was introduced when multidetector-row computed tomography (MDCT) was already spread worldwide. The expanding use of CTU has resulted in an increase in both the frequency of procedures and patient radiation exposure (1, 2). Considering that the urinary tract is a high-contrast structure during the excretory phase (EP), substantial dose reduction may be possible while maintaining diagnostic efficiency (3).
In modern MDCT systems, several approaches have been used in an effort to minimize the radiation dose, such as automatic tube current modulation (ATCM), lowering the tube current, and using a noise reduction filter (NRF). The ATCM technique can achieve CT dose index (CTDI) reductions of 40–60% without affecting image quality and is now routinely used on most CT systems (4, 5). The non-linear NRF technique can reduce image noise up to 30–60% of its original level, and its use has recently been reported in several anatomical regions (6–9). To the best of our knowledge, low-dose EP images using ATCM and a postprocessing NRF (pNRF) have not been reported in the medical literature. In addition, there are no data regarding the appearance of the urinary bladder in thin-slice EP images, including CT attenuation and the size of the urinary bladder.
In our institution, abdominal-pelvic CT using a noise index (NI) of 9.88 in a slice thickness of 5 mm was initially conducted in patients with confirmed or suspected urothelial malignancies; therefore, a dose reduction protocol was required to avoid high dose exposure. Meanwhile, our preliminary study (not published) showed that image noise in the area containing the anterior pelvic fat was approximately 12.0 in this protocol, and it was shown that diagnostically acceptable images could be obtained using abdominal – pelvic CT by employing an NI of 12 (10). Consequently, it can be assumed that EP images using an NI of 20 and pNRF would be similar to those using an NI of 12 without a filter due to the 40% noise reduction effect of the NRF; therefore, we attempted to qualitatively and quantitatively evaluate EP images using a high NI and a pNRF and compare them with those using a low NI.
Material and Methods
The human research committee of our institutional review board approved our study protocol which involved a retrospective evaluation of CT scans, radiation doses, and medical records, and the requirement to obtain informed consent was waived.
In a preliminary study, we found that the radiation dose delivered during abdominal-pelvic CT using a low NI, which was only applied to patients who were highly suspected of suffering from urothelial malignancy, was 16–17 mGy per phase. Our ethical committee approved the CTU protocol because they understood that the lower noise would allow small malignancies in the urinary tracts to be detected on reconstructed thin slice images and that an increased dose was justified in patients who were at increased risk of malignancy. Moreover, the mean dose delivered during nephrogenic phase (NP) and EP scans did not exceed 35 mGy, which is the radiation dose recommended in the EUR 16262 European Guidelines On Quality Criteria For Computed Tomography (11).
Phantom and clinical studies were performed using a 32-MDCT system (Lightspeed VCT-select; GE Healthcare, Milwaukee, WI, USA) with a standard reconstruction algorithm. All images were scanned using the xyz-axes modulation technique (Smart mA; GE Healthcare), and a non-linear 3D-NRF (Neuro3D filter; GE Healthcare) in high level mode (greatest noise reduction of the three possible modes) was used for the high NI images.
Phantom experiment
Before the clinical study, the effect of noise reduction with a pNRF was analyzed. A water phantom constructed of Plexiglas measuring 20 × 20 × 32 cm in a cylindrical column (helical CT Phantom HT type; Kyoto Kagaku, Kyoto, Japan) was scanned four times. The scanning parameters were as follows: collimation 32 × 1.25 mm; rotation time 0.4 s/rot; pitch 1.375; tube voltage 120 kVp; minimal and maximal tube current thresholds of 10 and 650mA, respectively; a matrix of 512 × 512; and a field of view (FOV) of 35 cm. The NI value was defined for a slice thickness of 5 mm, and reconstructed images of 1.25 mm in width were evaluated. Scans were performed at an NI of 9.88 (low NI; range: 45–78 mA), followed by an NI of 20 (high NI). Each image was reconstructed at a slice thickness of 1.25 mm, and the images using an NI of 20 were immediately postprocessed using the NRF.
Image analysis
The standard deviation (SD) of attenuation (Hounsfield units [HU]) was measured as an index of image noise in each image, using circular regions of interest (ROIs) of 300 mm2 (Table 1). ROI measurements were taken in the center and four peripheral regions (3, 6, 9, and 12 o'clock positions) within the phantom, and in three regions of the air outside the phantom (above, right side, and left side). After the mean SD value had been calculated, the noise reduction rate (%) was calculated as follows: [1–(SD on images using an NI of 20 with pNRF) / (SD on images using an NI of 20) ] ×100 (Table 1). Scanner-generated CTDIvol values (mGy) were recorded, and the mean CTDIvol value was calculated (Table 1). The percentage radiation exposure reduction rate (%) was calculated as follows: [1–(CTDIvol on images using an NI of 20 and pNRF) / (mean CTDIvol on images using an NI of 20) ] ×100 (Table 1).
Image noise and noise reduction with non-linear noise reduction filter in a phantom study
NI = noise index
*Percentage reduction rate (%) of image noise was calculated as follows: [1–(SD on images set at NI of 20 with postprocessed NRF) / (SD on images setting NI of 20) ] ×100
†Percentage reduction rate (%) of radiation exposure was calculated as follows: [1–(mean CTDIvol on images set at NI of 20) / (mean CTDIvol on images set at NI of 9.88) ] ×100
Patients
Sixty-nine patients who were considered to be at high risk of urothelial cancer underwent CT scans from December 2005 to April 2008. The CT examinations were divided into two protocols according to when they were performed. The first set of CT examinations, which consisted of EP scans using an NI of 9.88, was performed in 34 patients between December 2005 and March 2007 (hereafter referred to as group A). The second set of CT examinations, which consisted of EP scans using an NI of 20 and pNRF, was performed in 35 patients between April 2007 and April 2008 (hereafter referred to as group B). Of the 69 patients, five patients in group A and 3 patients in group B were excluded, as they had hip prostheses, and one patient in group B who could not undergo the urine-mixing procedure was also excluded. The remaining 60 patients (51 men and nine women, age range 48–85 years, mean age ± SD 69.8 ± 8.6 years) were included in the study. Transurethral resection of bladder tumors (TUR-B) was performed in 58 patients, and total cystectomy was performed in two patients within 60 days of the CT examinations (range 4–60 days, mean period ± SD 22.6 ± 26 days). Forty-six patients (21 in group A and 25 in group B) had pathologically confirmed bladder tumor (BT) that had been classified as transitional cell carcinoma (TCC), six had carcinoma in situ, three had metaplasia, three had inflammation, one had papilloma, and one had an adenoma. The radiological and clinical characteristics of the BT were polypoid or sessile-shaped lesions in the bladder lumen, and focal wall thickening (12, 13). Of the 46 patients with BT, 33 (15 in group A and 18 in group B) had at least one polypoid or sessile-shaped superficial BT with a short-axis diameter greater than 10 mm. The other nine patients had superficial BTs only with a short axis diameter of less than 10 mm, and four patients had invasive BT with focal wall thickening.
All patients had undergone cystoscopy before CTU examinations due to a high suspicion of BTs. Macrohematuria with urinary cytologic abnormalities was the initial symptom in 37 of the 60 patients (62%). Cystoscopic abnormalities were found in the other 23 patients (38%) who had received intravesical immunotherapy or chemotherapy after TUR-B. No patients had previous allergic reaction to iodinated contrast material (CM), renal insufficiency (serum creatinine level of >1.3 mg/dl), or a previous history of asthma. We recorded the patients' BT status (initial or recurrent), whether they had pathologically confirmed TCC, and the number of polypoid or sessile-shaped BTs with a short-axis diameter greater than 10 mm (Table 2). The scanner-generated CTDIvol values (mGy) during the NP and EP and the maximal tube current (mA) in the bladder were recorded (Table 2). In addition, age (years), sex (men/women), body weight (kg), height (cm), and FOV (cm) were recorded as characteristic parameters (Table 2). All data are shown as mean values.
Patient characteristics with highly suspicion of bladder tumors
Values are mean ± SD in the parentheses
BT = bladder tumor; – = not applicable
CT urography imaging protocol
Our CTU protocol consisted of NP and EP imaging. The table speed was set at 137.5 mm/s in both phases and was performed in the same fashion as the phantom study, including tube voltage, matrix size, and section thickness. The other acquisition parameters were as follows: a minimum tube current of 100 mA and a maximum tube current of 650 mA, and the FOV depended on the patient's physique (range 27.1–45.3 cm).
All patients were free to drink water until their arrival at our department, after which they were instructed not to void and were asked to ingest 300 mL water at least 20 min before the scan. After anterior and lateral localizer images had been obtained, NP images from the top of the diaphragm to the lower end of the pubic symphysis were scanned in the supine position 90 s after 100 mL iodinated (300 mg I/mL) CM (Iohexol; Daiichi-Sankyo, Tokyo, Japan, Iomeprol; Ezai, Tokyo, Japan, Iopamidol; Bayer Healthcare, Osaka, Japan) had been administered at a rate of 2.0 mL/s into an antecubital vein using a power injector. No saline flush was used. EP images from the top of the kidneys to the lower end of the pubic symphysis were scanned in the prone position 480 s after the CM injection (14).
On the basis of the assumption that a 45-degree tilting procedure might mix the unopacified urine and contrast-enhanced urine homogenously within the bladder, technologists held the patient's waist firmly and placed it in a right oblique position (an approximately 45-degree angle) relative to the fulcrum of the left iliac bone, and then in a left oblique position relative to the fulcrum of the right iliac bone. Clockwise and anticlockwise tilting procedures were performed on the CT table.
After localizer images of the EP images had been obtained, the 45-degree tilting procedure was performed more than three times for at least 30 s prior to the start of the EP scan with the patient's position remaining unchanged.
Image analysis
Source axial images were loaded onto a separate workstation (Advantage workstation 4.2; GE Healthcare) and were displayed in an angiographic window with a window/level setting of 600/200 HU (15). To analyze the entire urinary bladder on the workstation, axial images with a slice thickness of 1.25 mm were usually used. Sagittal and coronal multiplanar reconstruction (MPR) images with a slice thickness of 0.6–0.8 mm were utilized to measure the maximal size of the opacified bladder.
Quantitative assessment
Quantitative analysis was performed by one experienced radiologist (YK). Axial images were used for the ROI measurements. CT attenuation (HU) in the urinary bladder was measured in the center at different three levels, the upper, middle, and lower thirds of the bladder, using circular ROI of 500–800 mm2. CT attenuation and image noise in the anterior pelvic fat (the subcutaneous fat located at the anterior pelvic wall) were also measured, using ellipsoidal ROI of 300–500 mm2 at the same three levels (Fig. 1). After the mean attenuation and image noise had been calculated for each measurement, the contrast-to-noise ratio (CNR) of the urinary bladder with respect to the anterior pelvic fat was calculated with the following formula: [(attenuation of the urinary bladder) – (attenuation of the anterior pelvic fat)] / image noise of the anterior pelvic fat (3, 10).
Multiplanar reconstructed images using an NI of 9.88 in a 65-year-old man of 68 kg and 165 cm, who was suspected of having a BT. (a) A coronal image shows a polypoid BT on the right side of the bladder neck as a defect (arrow) and three different lines of the bladder. (b) A sagittal image also shows three different lines, the upper, middle, and lower thirds of the urinary bladder. (c) An axial image demonstrates the level of the lower margin of the acetabulum in the upper third. (d) An axial image demonstrates the vesicoureteral junction is present in the middle third of the bladder. A streak artifact due to beam hardening is seen within the bladder. (e) An axial image demonstrates the level of the upper margin of the pubic symphysis in the lower third of the bladder. The mean attenuation of the urinary bladder was 162.8 HU
In addition, the maximal size of the opacified bladder was measured using the longitudinal, transverse, and depth diameters (mm), using axial, sagittal, and coronal MPR images (Fig. 2). The short axis diameter (mm) of the layering contrast within the bladder and the unopacified urine in the bladder were also measured, using axial images on the fixed window/level setting (Fig. 3). The mean value of each diameter measurement was then calculated.
Diameter of the opacified bladder on EP images using NI of 9.88 in a 55-year-old man of 77 kg and 171 cm. (a) A coronal image shows a longitudinal diameter of the opacified bladder of 90 mm. (b) A sagittal image shows a depth diameter of 71 mm. (c) An axial image shows a transverse diameter of 84 mm
Measurement of the short-axis diameters of the layering contrast and unopacified urine. (a) An axial EP image using an NI of 9.88 in a 72 year-old-man of 60 kg and 158 cm showed that the short-axis diameter (arrows) of the layering contrast within the bladder was 3.5 mm. (b) An axial EP image using an NI of 20 and pNRF in a 77 year-old-man of 65 kg and 179 cm showed that the short-axis diameter (arrows) of the unopacified urine within the bladder was 18 mm
Qualitative analysis
Images of the urinary bladder were analyzed in random order and by the consensus of two radiologists (TH and YO), who graded image noise, sharpness, streak artifacts, homogeneity, and conspicuity of polypoid or sessile-shaped BTs with a short-axis diameter greater than 10 mm using a 3-point scale (4, 6, 10, 16). Image noise, which was defined as ‘graininess’, was categorized using the following 3-point scale: 3, fully acceptable; 2, probably acceptable; and 1, unacceptable due to severe noise. Sharpness to identify surface irregularity to delineate the urinary bladder contour was graded using the following 3-point scale: 3, acceptable; 2, probably acceptable; and 1, unacceptable. Streak artifacts due to beam-hardening were graded using the following 3-point scale: 3, very minimal; 2, moderate; and 1, severe. The homogeneity of bladder urine was graded using the following 3-point scale: 3, homogenous; 2, intermediate homogeneity; and 1, inhomogeneous. In addition, the EP images from 33 patients (15 in group A and 18 in group B) who had one or more polypoid or sessile-shaped BT with a short-axis diameter greater than 10 mm (17) were subjected to grading of tumor conspicuity to identify clear filling defects (Fig. 4) using the following 3-point scale: 3, excellent; 2, acceptable; and 1, poor.
Three-point scale used for the qualitative assessment of tumor conspicuity on EP images using an NI of 20 and pNRF in patients who had more than one polypoid or sessile-shaped BTs (arrows) with a short-axis diameter greater than 10 mm. (a) Score 1, poor conspicuity, different to identify BT as a filling defect (arrow) within the bladder; (b) score 2, acceptable conspicuity, possible to identify BT as a filling defect within the bladder; and (c) score 3, excellent conspicuity, easy to identify BT as a filling defect within the bladder
Statistical analysis
Statistical analysis was performed using SPSS 14.0 (Chicago, IL, USA). Patient characteristics were compared using the unpaired Student's t-test. Sex distribution and patients with BTs were analyzed with the chi-square test. CT attenuation and image noise in the bladder and anterior pelvic fat, the maximal size of the opacified bladder, and the urinary diameter were evaluated using the unpaired Student's t-test. The Mann-Whitney U test was used for all qualitative assessment. Probability values of less than 0.05 were considered significant. Inter-observer variability in each qualitative evaluation was assessed with the kappa coefficient of concordance to measure the degree of agreement between the two radiologists. A kappa value lower than 0.20 indicated poor agreement; 0.21–0.40, fair; 0.41–0.60, moderate; 0.61–0.80, good; and 0.81–1.0, excellent.
Results
Phantom study
Table 1 showed that the dual NI protocol led to a 25% dose reduction. The mean tube current on the images using an NI of 20 and NRF (60mA) showed a 75% dose reduction compared to those using an NI of 9.88 (243 mA). Up to 40% noise reduction was achieved with the pNRF in each position. Although image noise in the center of the phantom corresponded to the SD, the noise at the periphery and outside of the phantom was smaller than at the center. In the images using an NI of 20 and pNRF, the image noise, apart from that at the center of the phantom, was similar to that using an NI of 9.88.
Patient study
Table 2 showed that patient characteristics were not significantly different between the two groups except for the radiation dose on EP. CTDIvol value on EP demonstrated 73% reduction in group B (4.6 ± 1.1 mGy), compared with group A (16.9 ± 3.2 mGy). In group B, the maximal tube current on the bladder (212 mA) could not be achieved with a high mA.
In the quantitative assessment, the CNR value of the urinary bladder with respect to the anterior pelvic fat was not significantly different (P = 0.75) between group A (16.1 ± 5.3) and group B (16.6 ± 7.6). In addition, there was no significant difference between the two groups in the size of the urinary bladder, the short-axis diameter of layering contrast, and the short-axis diameter of unopacified urine within the bladder. Short-axis diameter of BTs greater than 10mm did not show difference between the groups (Table 3).
Quantitative evaluation of the urinary bladder in the pelvis
Values are mean ± SD in the parentheses
BT = bladder tumor
Regarding the qualitative evaluation, sharpness, streak artifacts, urinary opacification, and tumor conspicuity for polypoid or sessile-shaped BTs with a short-axis diameter greater than 10 mm were not significantly different between groups (Table 4). Image noise in group A was superior (P < 0.01) to in group B.
Qualitative assessment of the urinary bladder in the pelvis
BT = bladder tumor
The kappa coefficients for sharpness, streak artifacts, homogeneity, and the conspicuity of polypoid or sessile-shaped BTs with a short-axis diameter greater than 10 mm were 0.66–0.75, indicating good inter-observer agreement in the qualitative assessments.
Discussion
This study showed that an NRF enabled approximately 40% noise reduction in a phantom study, and EP images using an NI of 20 and pNRF were quantitatively and qualitatively comparable to those using an Nl of 9.88, except with regard to image noise.
As the NI is dependent on the SD for a reconstructed slice thickness of 5 mm, reducing the slice thickness from 5 mm to 1.25 mm causes the X-ray intensity to fall by a factor of four. Consequently, the noise increases by the square root of four, from 100% to 200%. In addition, image noise is inversely proportional to the square root of the radiation dose, and therefore, doubling the NI should result in a 75% reduction in radiation exposure. These basic concepts were confirmed in the phantom study and the clinical EP imaging study by the reduction of the CTDIvol value.
Regarding the quantitative assessment in the clinical study, the mean CNR value for 16.4 will allow the successful identification of wall abnormalities, as Coppernrath et al. (3) described in a ureter phantom composed of CM-filled tubes that images with quality grade of ‘very good’ showed a CNR ranging from 10 to 20. The CNR of the urinary bladder with respect to the anterior pelvic fat did not show a significant difference between the images using a high NI and pNRF and those using a low NI, although the image noise was different in the phantom study. This was mainly due to the fact that the image noise caused by the anterior pelvic fat was not different in either group, and it can be considered that the measurement of background noise in the human body, which contains non-homogeneous and non-circular object, differs from that in the phantom which contains homogenous and circular object. In addition, the percentage tube current reduction ratio (%) between the two groups was different between the phantom and clinical studies. In the phantom study, the maximal tube current with an NI of 20 with pNRF was approximately 25% for an NI of 9.88; however, in the clinical study, group B was 35.6% of that observed in group A. We assumed that the maximal pelvic CT tube current in group A (594 mA) might reach a high mA level (650 mA). With respect to ATCM, the NI concept refers to the standard deviation of pixel values in the center of a water phantom. The noise on CT scans is dominated by the projections in which attenuation is highest, such as the iliac bone on pelvic CT in the lateral position. If the mean tube current parameter is less than the tube current required to generate images with the required NI, a photon starvation situation occurs in the pelvis (18). No other quantitative analyses produced a significant difference between the two groups. The results of the urinary homogeneity assessment, including its noise, and the diameter of laying contrast and unopacified urine within the bladder, indicated that the 45-degree tilting procedure for mixing the urine and CM achieved homogeneity.
In the qualitative assessment, a visual difference in graininess was detected, although CNR with respect to the anterior pelvic fat did not differ between groups. We presumed that this was due to the increased NI setting (reduced tube current), and approximately 75% dose reduction could not revert to the original graininess, even when the pNRF technique was used. Furthermore, the adaptive statistical iterative reconstruction (ASIR) algorithm, which reduces the quantum noise, would be an effective tool for improving CTU image quality (19, 20).
High radiation exposure is one of the major drawbacks of CTU. The European Society of Uroradiology (ESUR) suggested that the CTDIvol on EP images is 5–6 mGy in average-sized patients (60–80 kg), and the mean dose (4.6 mGy) on images using a high NI in this study corresponded to the standard dose of ESUR (2). The radiation dose (mean dose of 16.6 mGy in 60 patients) delivered during the NP in both groups and the EP in group A was much higher than the ESUR recommended dose for patients at high risk of malignancy (9–12 mGy). With regard to our justification for this high radiation dose, this protocol was only applied to patients at high risk of malignancy. In addition, although it was reported that the diagnostic capability of CTU was suboptimal (21), our assumption that lowering the radiation dose and using thin-slice reconstruction in CT examinations would increase image noise, and might reduce the likelihood of detecting small BTs was accepted at our institution.
In the present study, the ranges and mean body weights were smaller than those of people in North America and Europe. When heavier patients are scanned using the low NI protocol, the total radiation dose may increase, because the mean tube current may increase in the upper abdomen despite the remaining mean tube current in the pelvis with a constant maximum mA.
The present study has several limitations. First, the two groups were not randomly selected and this was a retrospective study, so further studies are required to validate the present results in a prospective and random manner. Second, only polypoid and sessile-shaped BT with a short-axis diameter greater than 10 mm were investigated for tumor conspicuity (16); therefore, the assessment of small lesions less than 10 mm is necessary. Third, 300 mL water was lower than recommended amounts, and the EP delay of 8 min without furosemide was shorter than that recommended in the ESUR guidelines (1, 2); therefore, it is necessary to ingest more water, and to perform the scan after a longer delay.
In conclusion, EP images using a high NI and pNRF were quantitatively and qualitatively comparable to those using a low NI, except with regard to image noise.