Obesity represents one of the largest socioeconomic and health care challenges in Western countries. Currently, more that 35% of U.S. adults are obese, and the prevalence is increasing [
1–
4]. Obesity is associated with substantial morbidity and mortality, mainly regarding cardiovascular diseases and cancer [
5,
6]. Thus, obese patients have an increased demand for medical care and along with this, for diagnostic imaging [
5,
7,
8].
In obese patients, x-ray–based imaging, such as CT, is hampered by reduced image quality, which is mainly caused by the photon starvation effect. The attenuation of photons increases exponentially with the distance that must be passed through mass [
9]. This results in increased noise and degradation of image quality. Thus, reduction of radiation dose for obese patients often is not possible [
10].
Standard CT scanners are equipped with detectors built in a discrete circuit (DC). X-rays are converted to photons that are converted to electric current by photodiodes. This analog signal is fed to analog-to-digital converters (ADCs) on a separate board. This detector setup has the major disadvantage of a longer path length from the photodiodes to the ADCs. A new generation of detectors combines the photodiodes and the ADCs on a single board into one application-specific integrated circuit (IC) detector. This reduces the length of the transmission of the analog signal and thus has the potential of reducing power consumption, heat dissipation, and electronic noise, which translate into reduced image noise [
11]. This could be particularly evident in overweight and obese patients because electronic noise becomes more and more dominant with increases in the amount of mass x-rays have to pass through [
12].
The purpose of our study was to assess the effect of the IC detector on image quality and noise in abdominal CT. We first determined the impact of the IC detector in an ex vivo study using various-sized abdominal phantoms. Second, we assessed images in patients undergoing abdominal CT with the IC detector and compared those with images acquired in the same patients undergoing abdominal CT on the same scanner with the same protocol settings but with conventional DC detector technology.
Materials and Methods
Abdominal Phantom
We used a commercially available anthropomorphic abdominal phantom placed in an inlay with different density cylinders (QRM-DEP-002, QRM). The phantom contains nine pairs of cylinders with different attenuation values ranging from fat (−100 HU) and tissue-equivalent (57 HU) to various defined fat and tissue mixtures with iodinated contrast material (ranging from −52 to 209 HU). The cross-sectional diameter of the abdominal phantom was 200 × 300 mm.
To simulate an overweight and obese patient, the phantom was modified by adding different circumferential layers. Two tissue-equivalent extension rings of different sizes but similar thickness (40 HU) and one fat equivalent ring (−60 HU) were available. Using these extension rings, a total of four different patient sizes were simulated (
Fig. 1): the phantom without extensions, hereafter called “small-sized” (300-mm maximal diameter); the phantom with the smaller tissue-equivalent extension ring, hereafter called “medium-sized” (350-mm maximal diameter); the phantom with the bigger tissue-equivalent extension ring, hereafter called “large-sized” (400-mm maximal diameter); and the phantom with the fat extension over both of the 5-cm tissue-equivalent rings, hereafter called “x-large-sized” (600-mm maximal diameter).
CT of the Phantom
The CT data were acquired with a second-generation dual-source 128-MDCT scanner (Somatom Definition Flash, Siemens Healthcare). The phantom was scanned on a CT scanner equipped with the standard detector and on a similar CT scanner equipped with the IC detector (Stellar, Siemens Healthcare).
Scanning parameters were the same with both detectors: slice acquisition was 2 × 0.6 × 64, by means of a z-flying focal spot; gantry rotation, 0.5 seconds; tube voltage, 120 kV; tube current–time product, 150 effective mAs for full dose (volume CT dose index [CTDIvol], 10 mGy); and 75 effective mAs (CTDIvol, 5 mGy) for half-dose scans. For standardization purposes and because of the cylindric nature of the phantom, no automated attenuation-based tube current modulation was used. The reconstruction FOV was set at 500 mm with a pixel matrix of 512 × 512.
All CT datasets were reconstructed with a slice thickness of 2 mm and an increment of 1.6 mm. Images were reconstructed with filtered back projection (FBP), using a medium soft-tissue convolution kernel (B30f). Thus, a total of 16 datasets were available for further analysis. Images were analyzed on workstations using our PACS (IMPAX 6.4.0, Agfa Healthcare).
Subjective Image Quality Analysis of Phantoms
Two independent readers (reader 1, hereafter referred to as “R1”; and reader 2, hereafter referred to as “R2”; each with 3 years of experience in radiology), who were blinded to the detector used, the reconstruction algorithm used, and each other's results, evaluated the images. Window settings were fixed for both readers (window width, 550 HU; level, 50 HU), corresponding to our standard abdominal soft-tissue window setting.
Readers evaluated overall image quality on a 5-point Likert scale: score 1 = bad, no diagnosis possible; score 2 = poor, diagnostic confidence substantially reduced; score 3 = moderate, but sufficient for diagnosis; score 4 = good; and score 5 = excellent [
13].
In addition, readers evaluated the various cylinders regarding sharpness of the cylinder contours on a 5-point Likert scale: score 1 = severely blurred or unsharp, very poor, nondiagnostic; score 2 = noticeable blur or unsharpness, poorly defined edges; score 3 = moderate, slightly blurred or unsharp; score 4 = good, mildly unsharp edges; and score 5 = excellent [
13].
Objective Image Quality Analysis of the Phantoms
Attenuation values were measured in the cylinders by R1, who manually drew a circular region of interest (ROI) at a predefined size of 47.9 mm2 on corresponding slices for scans acquired with the different detectors showing the middle set of inlays. To minimize errors, measurements were performed on three consecutive transverse slices, and the mean of measurements was taken. The SD of the attenuation was taken as the measure of image noise.
Patient Population
Twenty consecutively scanned patients (eight women; mean age, 62 ± 9 years; age range, 42–71 years and 12 men; mean age, 62 ± 12 years; age range, 32–74 years) were included in the second part of the study. Patients were included if they underwent a clinically indicated contrast-enhanced abdominopelvic CT study and if a previously performed contrast-enhanced abdominopelvic CT study was available that was obtained on the same CT scanner using the same protocol settings but with the CT scanner equipped with a conventional DC detector. The contrast media injection protocols were similar for both studies. The mean time interval between CT studies with the DC and the IC detector was 15 months. Indications for abdominopelvic CT included staging in oncology (
n = 17) and follow-up after vascular intervention or surgery (
n = 3). Patient demographics are listed in
Table 1.
The study had institutional review board and local ethics committee approval. Written informed consent was waived. All CT studies were clinically indicated, and no CT studies were performed merely for the purpose of this study.
CT in Patients
Imaging in patients was performed using a dual-source 128-MDCT Somatom Definition Flash equipped with a conventional DC detector for scans acquired before March 2012 and a Stellar scanner equipped with the IC detector for scans acquired after March 2012. Scanning parameters were slice acquisition, 2 × 0.6 × 64 by means of a z-flying focal spot; gantry rotation, 0.5 seconds; tube voltage, 120 kVp; reference tube current-time product, and 150 reference mAs with automated attenuation-based tube current modulation (Care-Dose4D, Siemens Healthcare).
All images were reconstructed using FBP and a medium tissue convolution kernel (B30f) with a slice thickness of 2.0 mm and an increment of 1.6 mm. Images in patients with iterative reconstruction were not included in this study because iterative reconstruction was not available at the time of the initial CT studies with the DC detector at our institution. Images were analyzed on the same workstations described for the phantom study.
Subjective Image Quality Analysis in Patients
Both readers (R1 and R2) evaluated the images concerning overall image quality using a 5-point Likert scale: score 1 = bad, no diagnosis possible; score 2 = poor, diagnostic confidence substantially reduced; score 3 = moderate but sufficient for diagnosis; score 4 = good; and score 5 = excellent image quality [
1]. Window settings were fixed for both readers (window width, 550 HU; level, 50 HU), similar to the phantom study.
Objective Image Quality Analysis in Patients
R1 measured the maximum transverse abdominal diameter on a slice at the intermediate portion of the kidneys. Image noise was defined as the SD of CT attenuation in a circular ROI with an average size of 2 cm2 placed in the subcutaneous fat at the height of the intermediate portion of the kidneys. Additionally, the effective mAs of each scan was noted from the patient protocol.
Statistical Analysis
Continuous variables are expressed as mean ± SD and tested for normality using the Kolmogorov-Smirnoff test. The interreader agreement was assessed by calculating the intraclass correlation coefficient (ICC) [
14]. All comparisons were made between the DC and the IC detector for different size phantoms for reconstruction algorithms and high and low mAs settings resulting in 16 pairs. The Wilcoxon signed rank test was used to test for significant differences regarding subjective image quality. Noise, abdominal diameter, and effective mAs between datasets acquired with the DC and the IC detector were compared using the Student
t test for paired samples. The Pearson correlation coefficient was calculated to correlate abdominal diameter with image noise. Data analysis was performed using commercially available software (SPSS Statistics version 20, release 20.0.0, IBM). Statistical significance was inferred at a
p value below 0.05.
Discussion
Obesity represents an independent risk factor for various diseases, leading to the requirement of increased diagnostic imaging procedures in these patients [
7,
8]. One common way to achieve image quality in overweight and obese patients that is diagnostic and comparable to images in normal-weight patients is to increase the radiation dose [
9,
10]. Our combined in vivo and ex vivo study was aimed at an analysis of the effect of a recently introduced IC detector system on image quality, with a particular focus on abdominal CT imaging in overweight and obese patients. Both phantom experiments and in vivo patient data indicated a gradually increasing positive effect of the IC detector on image quality and noise with increasing body sizes, most pronounced in low-radiation-dose studies.
Reducing radiation dose results in increased image noise, which could hamper diagnostic confidence significantly [
15]. This is most probably due to the reduced electronic noise in the IC detector when compared with the DC detector. Electronic noise becomes a more significant source of image noise when fewer photons are arriving at the detector, the photon starvation effect. In overweight and obese patients, the photon starvation effect is aggravated because the photons have to penetrate more absorbing mass. This results in reduced signal at the detector, translating to a reduced analog signal. On standard detectors the ADC is installed on a separate board in a DC circuit fashion, which means the analog signals from the photodiodes have a longer path of transmission. The IC detector is reducing this transmission length by having the ADC installed on the same board. In this way, the board's configuration has the ADC and photodiodes attached to the backside of the ceramic scintillators. Thus, the analog signals fed from the photodiodes have a shorter path of transmission, which reduces the possibility of information loss and results in decreased electronic noise, as previously shown [
11].
In our patient study, readers found no significant improvement in subjective image quality but still rated some datasets acquired with the IC detector of higher overall image quality. Similar to the phantom study, we found a gradual increase in noise reduction with increasing body size (determined by abdominal diameter measurements). We found a significant noise reduction of an average of 6% across all body sizes for the IC detector compared with the DC detector. This increase was 8% in obese patients. The discrepancy in the absolute extent of noise reduction between the phantom and the patient study might be explained by the use of automated tube current modulation applied in the patients. In addition, we did not include low radiation dose studies in patients. Thus, the highest amount of noise reduction, as seen in the phantom experiments, could not be demonstrated.
However, we decided to not use automated tube-current modulation in the phantom study. This was done to obtain standardized scans that would be more comparable in the phantom study and to exclude other reasons for a difference in image noise that could have occurred due to automated tube-current modulation.
Noise and body size correlated in the FBP images acquired with the DC detector. This indicates that, despite the use of automated attenuation-based tube current modulation, noise does not remain constant over all patient sizes [
16]. With the IC detector, correlation was weaker (but not significant), indicating the reduced dependence of image noise on body size when using the IC detector.
The following study limitations must be addressed. We could not show how the increased image quality in the phantom study could translate to an increase in diagnostic accuracy because the phantom used had a repetitive inlay configuration that was not designed to test for lesion detection. Furthermore, we did not include iterative reconstructions into our analyses. However, it might be assumed that the observed effect of the IC detector is additive to the known body size–independent effect of iterative reconstructions in terms of image noise [
17,
18].
In conclusion, our study indicates that use of the IC detector for abdominal CT is associated with increased image quality and reduced noise, an effect that is pronounced in over-weight and obese patients and in low-radiation-dose studies. Thus, the IC detector might be used to translate the improved image quality to a reduced radiation dose in overweight and obese patients.