Medical Physics and Informatics Original Research Duan et al. Electronic Noise in CT Detectors Medical Physics and Informatics Original Research Xinhui Duan 1 Jia Wang 1,2 Shuai Leng 1 ernhard Schmidt 3 Thomas llmendinger 3 Katharine Grant 3 Thomas Flohr 3 Cynthia H. McCollough 1 Duan X, Wang J, Leng S, et al. Keywords: artifacts, electronics, noise, radiation dosage, x-ray CT DOI:1.2214/JR.12.1234 Received October 29, 212; accepted after revision December 3, 212. 1 Department of Radiology, Mayo Clinic, CT Clinical Innovation Center, 2 First St SW, East 2 Mayo ldg, Rochester, MN 5595. ddress correspondence to C. H. McCollough (mccollough.cynthia@mayo.edu). 2 Stanford University, Stanford, C. 3 Siemens Healthcare, Forchheim, Germany. WE This is a web exclusive article. JR 213; 21:W626 W632 361 83X/13/214 W626 merican Roentgen Ray Society Electronic Noise in CT Detectors: Impact on Image Noise and rtifacts OJECTIVE. The objective of our study was to evaluate in phantoms the differences in CT image noise and artifact level between two types of commercial CT detectors: one with distributed electronics (conventional) and one with integrated electronics intended to decrease system electronic noise. MTERILS ND METHODS. Cylindric water phantoms of 2, 3, and 4 cm in diameter were scanned using two CT scanners, one equipped with integrated detector electronics and one with distributed detector electronics. ll other scanning parameters were identical. Scans were acquired at four tube potentials and 1 tube currents. Semianthropomorphic phantoms were scanned to mimic the shoulder and abdominal regions. Images of two patients were also selected to show the clinical values of the integrated detector. RESULTS. Reduction of image noise with the integrated detector depended on phantom size, tube potential, and tube current. Scans that had low detected signal had the greatest reductions in noise, up to 4% for a 3-cm phantom scanned using 8 kv. This noise reduction translated into up to 5% in dose reduction to achieve equivalent image noise. Streak artifacts through regions of high attenuation were reduced by up to 45% on scans obtained using the integrated detector. Patient images also showed superior image quality for the integrated detector. CONCLUSION. For the same applied radiation level, the use of integrated electronics in a CT detector showed a substantially reduced level of electronic noise, resulting in reductions in image noise and artifacts, compared with detectors having distributed electronics. s a result of ongoing developments in CT technology, the number of CT examinations performed each year has continued to increase [1]. Radiation from CT examinations contributes a major proportion of patient and population doses, and the related potential cancer risk is of great concern [2 4]. One of the possible ways to reduce patient dose is to lower the technical parameters that govern scanner radiation output, such as x-ray tube voltage and current. However, as radiation dose levels are decreased, image noise increases, sometimes quite dramatically. The source of noise is mainly from the quantum noise properties of x-ray photons and the electronic noise of the detection system. X-ray photons carry useful information about patient anatomy, and the noise associated with the random nature of photon interactions (i.e., quantum noise) is related to the number of photons detected. However, electronic noise originates from the x-ray detection system; it is unrelated to the number of photons detected and does not carry any diagnostic information. nalog electronic circuits in the detection system are the main source of electronic noise. Once the analog signal is converted to a digital signal, it becomes relatively immune to sources of electronic noise (Fig. 1). In modern CT systems, electronic noise usually has a negligible effect for protocols using typical dose levels and for normal-size patients, but for low-dose protocols or obese patients, electronic noise becomes more important because the detected signal may become comparable to the electronic noise level [5]. s CT practice moves toward lower-dose scanning and as the number of obese patients continues to increase, electronic noise could have a significant impact on image quality and could become a primary constraint to the ability to reduce patient dose. If electronic noise can be reduced in an x-ray detection system, there is potential for even further patient dose reduction. Modern CT systems are equipped with solid-state detectors. Each detector cell consists W626 JR:21, October 213
Electronic Noise in CT Detectors of a radiation-sensitive solid-state material (e.g., cadmium tungstate, gadolinium oxide, or gadolinium oxysulfide) that converts the absorbed x-rays into visible light. The light is then detected by an attached silicon photodiode. In a traditional detector design, the small analog electrical current from the photodiode is amplified and converted into a digital signal on an external board, requiring analog connections between the detector cells and the electronic circuit components on the external board. Recently, detectors featuring fully integrated electronics were introduced to a commercial CT system. In these detectors, the photodiode, the preamplifiers, and the analog-to-digital converters are integrated into the same silicon chip that is attached to the scintillation ceramic, obviating analog connections. The integration of the electronics with the detector element reduces the time during which the signal is in analog form, thereby reducing the amount of electronic noise that can be added to the signal [6]. These two types of detectors are referred to as conventional and integrated detectors, respectively. For this study, we evaluated in phantoms the differences in CT image noise and artifact level between commercial CT detectors: a conventional detector with distributed electronics and an integrated detection system (Fig. 2). Integrated detectors from only a single manufacturer were evaluated because, to our knowledge, a similar design is not commercially available on other systems. Materials and Methods Phantom Evaluation Cylindric water phantoms of 2, 3, and 4 cm in diameter were scanned using two CT scanners: one equipped with an integrated detector (Stellar detector, Somatom Definition Flash, Siemens Healthcare) and the other with a conventional detector (Somatom Definition S+, Siemens Healthcare). oth detectors used the same scintillating material to convert x-rays to visible light photons. ll scanning parameters were otherwise identical. Scans were acquired at four tube voltages (8, 1, 12, and 14 kv), 1 tube currents (6 6 m), and a.5-second rotation time. Images were reconstructed with 1- and 5-mm image thicknesses and a medium smooth kernel (3). Two semianthropomorphic phantoms were also scanned to mimic the shoulder and abdominal regions with tube voltages of 8 and 12 kv and a quality reference tube current exposure time product of 24 ms. The images of cylindric water phantoms were evaluated for noise and the anthropomorphic phantoms were evaluated for artifacts. Image noise was quantified as the SD of CT numbers (STD) in uniform regions of interest (ROIs). rtifact level was quantified as the square root of the difference in STD 2 between regions containing streak artifacts (region 1) and adjacent artifact-free ROI (region 2) as follows: STD 2 1 STD 2 2. Noise power spectra were also computed [6]. Image noise versus tube current and noise reduction versus tube current were plotted to characterize the behavior of the detector systems and the results were compared with an ideal detector in which the signal depends only on detected x-ray photons. lthough an ideal detector does not exist in reality, it is often used as a reference against which to benchmark a practical system. The number of photons detected by an ideal detector generally follows a Poisson distribution, which means the variance (STD 2 ) equals the mean number of detected photons. ecause of the logarithm operation in the actual CT image reconstruction, the noise (STD) in CT images is inversely proportional to the square root of the mean photon number and thus is inversely proportional to the square root of the tube current. Therefore, when the signal amplitude from detected photons is much higher than the electronic noise, the curve of image noise versus tube current closely matches the inverse square root characteristics that is, the exponent on the fitted power equation equals.5. When the signal amplitude from electronic noise is of a magnitude comparable to that from the detected photons, the contributions to the signal from electronic noise become significant, and the curve of noise versus tube current becomes steeper in the low-tube-current (low-dose) region. Curves of noise versus tube current were fit to a power equation to determine the value of the exponent. Exponent values close to.5 behave very closely to an ideal detector. decrease in the exponent values from.5 indicates an increased role in the electronic noise contribution to the image. Demonstration of Clinical Value To show the clinical implications of the use of an integrated detector, we reviewed our image archive to identify patients who had been scanned on the evaluated scanner using the conventional detector and subsequently again after the installation of the integrated detectors. For inclusion, the clinical indications and protocol settings were required to be similar. This study was approved by the local institutional review board. Results Phantom Evaluation For cylindric phantoms, the reduction of image noise due to the use of integrated electronics depended on phantom size, tube potential, and tube current. Low-signal scans that is, low tube currents, low tube potentials, and large phantoms had larger noise reductions because electronic noise has the strongest effects in these situations. Figure 3 shows an example of a curve of image noise versus tube current. The amount that the radiation dose could be reduced from that used with a conventional detector to achieve a specific image noise level is shown in Figure 3. Depending on the acceptable noise level, the dose reduction ranged from 3% to 5% for this particular condition (8 kv and 3-cm phantom). Figure 4 shows the curves for noise reduction versus tube current for different tube potentials and phantom sizes for a 1-mm image thickness. The results for a 5-mm image thickness (data not shown) were essentially the same. In Figure 4, the noise level at 8 kv reaches a plateau at approximately 15 m. In Figure 4C, at 1 kv, the noise level decreases as the tube current is decreased from 2 to 1 m. These unexpected behaviors occur when the number of photons reaching the detector is below a minimum level set by the manufacturer. When this happens, image noise is no longer related to tube current in an exponential fashion. Streak artifacts through regions of high attenuation (e.g., through the shoulders) were greatly reduced for the integrated detector, especially in lower-dose scans. n example image of a shoulder phantom is given in Figure 5. The images reflect an artifact reduction of 45%. n example of the calculated noise power spectra is shown in Figure 6. The noise amplitude from the integrated detector was much lower than that from the conventional detector, which is consistent with image noise measurements. More important is the fact that the shape of the two spectra that is, the peak location was the same, which indicates that the noise texture in the images was not affected by the reduction in electronic noise. Demonstration of Clinical Value Two patients that met the inclusion criteria were identified. pediatric low-dose chest CT examination performed for oncology follow-up showed a subtle reduction in streak artifacts from the shoulder region (Figs. 7 and 7). The boy (age, 5 years) was scanned using a volume CT dose index (CTDI vol ) value of.62 mgy for both studies, which is an extremely low scanner output setting. CT examination of the chest, abdomen, and pelvis of a morbidly obese man (age, 37 JR:21, October 213 W627
Duan et al. years; weight, 26 kg) was performed. ecause the patient s size extended beyond the 5-cm-diameter scanning FOV, an extended- FOV reconstruction was performed to minimize truncation artifacts. The overall noise level was reduced for the examination performed using the integrated detector (Figs. 7C and 7D) even though a lower dose was used (CTDI vol for integrated detector examination vs conventional detector examination, 6.8 vs 7.3 mgy, respectively). Discussion The integrated detector resulted in substantially reduced levels of electronic noise, resulting in reductions in image noise and artifacts, compared with conventional detectors for the same scanner output level. lternatively, the integrated CT detector could be used to enable dose reduction for the same image noise level. For example, a lower dose was used to obtain Figure 7D than Figure 7C, but Figure 7D is less noisy. The most clinically significant impact is likely to be improved image quality in situations in which low photon counts are measured, such as in examinations using a low dose (Figs. 7 and 7) or examinations of large patients (Figs. 7C and 7D). In the 5-year-old child (Figs. 7 and 7), the reduction in electronic noise proved beneficial, even though the patient was small, because of the very low scanner output setting used. For a.62-mgy CTDI vol, the number of photons reaching the detector was very low even in a child low enough to be affected by electronic noise (Fig. 7E). The difference in noise behavior between the conventional and integrated detectors is small because the overall object attenuation is small. For the 37-year-old obese man (Figs. 7C and 7D), there was a large amount of attenuation, so the difference in the noise behavior between the conventional and integrated detectors is much larger and extends over the full range of tube current values (Fig. 7F). The benefit of the integrated detector is additive to other dose reduction techniques. For example, iterative reconstruction is being more widely used for CT dose reduction and is available from all the major CT scanner manufacturers. Some iterative reconstruction algorithms use detailed physical models of the CT scanner, including the sources of image noise and artifacts. This prior information is used in the image reconstruction process, which has the potential to improve image quality, such as increasing spatial resolution or reducing image noise. This modeling process is applicable to the integrated detector, just as it is with conventional detectors, so the benefits from these two techniques is additive. The integrated detector reduced the magnitude of noise and kept the shape of the noise power spectrum. Other noise reduction techniques, such as iterative reconstruction and image space denoising, can significantly change the noise texture in the image, and this change is reflected by a shift in the noise power spectrum to lower spatial frequencies. The different textures associated with iterative reconstruction methods are unfamiliar to radiologists and may cause difficulty in image interpretation. Many clinical applications may directly benefit from this integrated detector system for example, low-dose CT lung and colon screening in patients of all sizes. Streak artifacts commonly exist in these low-dose images through highly attenuating regions such as the shoulders or pelvis, respectively, which may deteriorate image quality and reduce diagnostic accuracy. The streak artifacts are caused by insufficient signal-to-noise ratios in these highattenuation projections. The integrated detector reduces electronic noise and thus improves the signal-to-noise ratio, reducing streak artifacts and improving image quality. For phantoms of the same size, we observed a larger difference in the noise behavior of the conventional and integrated detectors at lower tube potentials (8 and 1 kv) than at higher tube potentials (12 and 14 kv, data not shown). Thus, a greater benefit of the integrated detectors is expected when lower tube potential values are used to reduce patient dose [7]. There are limitations to this study. We did not evaluate the impact of image quality improvement or noise reduction on diagnostic accuracy or in patient examinations, but rather focused on quantitative characterization of the impact of reduced electronic noise levels, which in this work was achieved by integrating the detector electronics with the scintillating crystal. Complete clinical evaluations will be required to fully show the clinical value of this technical advancement. In conclusion, the use of integrated electronics in a CT detector resulted in substantially reduced levels of electronic noise, resulting in reductions in image noise and artifacts, compared with the use of detectors having distributed electronics. The use of an integrated CT detector has the potential to further reduce radiation dose to patients from CT examinations. References 1. IMV website. IMV 27 CT market summary report. www.imvinfo.com/user/documents/content_ documents/def_dis/28_6_12_6_45_57_76. pdf. Published 28. ccessed June 6, 213 2. errington de González, Darby S. Risk of cancer from diagnostic X-rays: estimates for the UK and 14 other countries. Lancet 24; 363:345 351 3. renner DJ, Hall EJ. Computed tomography: an increasing source of radiation exposure. N Engl J Med 27; 357:2277 2284 4. Mettler F Jr, Thomadsen R, hargavan M, et al. Medical radiation exposure in the U.S. in 26: preliminary results. Health Phys 28; 95:52 57 5. Massoumzadeh P, Don S, Hildebolt CF, ae KT, Whiting R. Validation of CT dose-reduction simulation. Med Phys 29; 36:174 189 6. oedeker KL, Cooper VN, McNitt-Gray MF. pplication of the noise power spectrum in modern diagnostic MDCT. Part I. Measurement of noise power spectra and noise equivalent quanta. Phys Med iol 27; 52:427 446 7. Yu L, Li H, Fletcher JG, McCollough CH. utomatic selection of tube potential for radiation dose reduction in CT: a general strategy. Med Phys 21; 37:234 243 (Figures appear on next page) W628 JR:21, October 213
Electronic Noise in CT Detectors mplitude (HU) mplitude (HU) 1..5..5 1..5 1..5 5 1 Time (U) 5 1 Time (U) 15 2 1 15 2 C mplitude (HU) mplitude (HU) 1..5..5 1...5.5 1 5 1 Time (U) 5 1 Time (U) 15 2 1 15 2 Fig. 1 Digital signals are more robust to noise than analog signals., Graph shows continuous analog signal., Graph shows continuous analog signal that has been contaminated by noise, thus changing amplitude of signal at each time point. C, Graph shows digital signal. D, Graph shows digital signal that has been contaminated by noise. However, because digital electronics allow only or 1 in dataset, information carried by noisy signal is exactly the same as information carried under noisefree conditions: 1,,1, in both signals. D Fig. 2 Photograph shows conventional CT detector module (top) (Somatom Definition S+, Siemens Healthcare) and integrated CT detector module (bottom) (Stellar detector, Somatom Definition Flash, Siemens Healthcare). Integrated detector directly couples photodiode with analog-to-digital convertor (DC), so amount of time that signal is in analog form is reduced, thereby reducing amount of electronic noise that can be added to signal. JR:21, October 213 W629
Duan et al. 3 25 Conventional detector Integrated detector Power (conventional detector) Power (integrated detector) Noise (SD of CT Number, HU) 2 15 1 5. Noise = 138 m.549 R 2 =.9986 Noise = 4952 m.724 R 2 =.9959 2 4 6 Dose Reduction (%) 55 5 45 4 35 3 25 2 4 6 8 1 12 14 Noise (SD of CT Number, HU) Fig. 3 Measured noise curves and calculated potential dose reduction., Graph shows curve for image noise versus tube current for 8-kV scan and 1-mm image thickness using 3-cm-diameter water phantom. Power-fit curves are shown, where exponential value of fit indicates degree to which detector acts like ideal detector (exponent =.5). Integrated detector has exponent of.549, whereas conventional detector has exponent of.724; these results indicate that integrated detector behaves much more like ideal detector than conventional detector., Calculated from data in, graph shows dose reduction that can be achieved using evaluated integrated detector relative to dose required by conventional detector to yield same image noise level. Having target noise level that is lower (less noisy) is achieved by using more photons. When signal amplitude from detecting more photons is relatively high, effect of electronic noise is negligible and hence dose reduction ability is not as high. Noise Reduction (%) Noise Reduction (%) 2 15 1 5 4 3 2 1 8 kv 1 kv 12 kv 1 2 3 4 5 6 1 kv 12 kv 14 kv 1 2 3 4 5 6 C Noise Reduction (%) 5 4 3 2 1 8 kv 1 kv 12 kv 14 kv 1 2 3 4 5 6 Fig. 4 Noise reduction for different kv, m, and phantom sizes. C, Graphs show curves of image noise reduction versus tube current for three phantoms with diameters of 2 (), 3 (), and 4 (C) cm and four tube potentials (8, 1, 12, and 14 kv). Some combinations of data acquired are not shown because signal level was too low to create reasonable image quality required for meaningful noise measurements. In, noise level at 8 kv reaches a plateau at approximately 15 m. In C, at 1 kv, the noise level decreases as the tube current is decreased from 2 to 1 m. These unexpected behaviors occur when number of photons reaching detector is below a minimum level set by manufacturer. When this happens, image noise is no longer related to tube current in an exponential fashion. W63 JR:21, October 213
Electronic Noise in CT Detectors Fig. 5 Images of semianthropomorphic phantom (lateral size = 36 cm, anteroposterior size = 16 cm). and, CT images (window center, 15 HU; window width, 65 HU) of semianthropomorphic shoulder phantom obtained from conventional () and integrated () detectors using 8 kv and 24 quality reference ms. Streak artifacts through region of high attenuation (e.g., through shoulders) are greatly reduced for integrated detector. In this example, image obtained with integrated detector reflects artifact reduction of 45%. Noise Power Spectra (HU 2 cm 2 1 4 ) 2.5 2. 1..5. Spatial Frequency (1/cm) Integrated detector Conventional detector 1 2 3 4 5 6 7 8 Fig. 6 Graph shows noise power spectra calculated from 8 kv, 6 m, and 1-mm slice thickness for 3-cm diameter phantom. Noise amplitude from integrated detector is much lower than that from conventional detector, which is consistent with image noise measurements. More importantly, shape of two spectra that is, peak location is the same, which indicates that noise texture in images was not affected by reduction in electronic noise. JR:21, October 213 W631
Duan et al. Noise (SD of the CT Number, HU) 7 6 5 4 3 2 1 2 4 6 8 C E Fig. 7 Sample patient images that benefit from use of integrated detector. and, CT images (window center, 4 HU; window width, 4 HU) of small patient (5-year-old boy) obtained using scanners with conventional detector () and integrated detector (). Volume CT dose index (CTDI vol ) is.62 mgy for both images. Reduction of electronic noise in image obtained using integrated detector is beneficial even though patient is small. C and D, CT images of large patient (37-year-old man; weight, 26 kg) obtained using scanners with conventional detector () and integrated detector (). CTDI vol is 7.3 mgy for image obtained with conventional detector and 6.8 mgy for image obtained with integrated detector. Overall noise level is reduced for examination performed using integrated detector even though CTDI vol is lower. E and F, Graphs illustrate behavior of noise and dose relation of conventional detector (blue line) and integrated detector (red line) for small (E) and large (F) patients. Noise (SD of the CT Number, HU) 4 35 3 25 2 15 1 5 2 4 6 8 D F W632 JR:21, October 213