Feasible Dose Reduction in Routine Chest Computed Tomography Maintaining Constant Image Quality Using the Last Three Scanner Generations: From Filtered Back Projection to Sinogram-affirmed Iterative Reconstruction and Impact of the Novel Fully Integrated Detector Design Minimizing Electronic Noise

INTRODUCTION

Since the introduction of computed tomography (CT) in clinical practice in the late 1970s, dose reduction has been a major concern for radiologists and their patients.[12] Over the last few decades, spiral CT, cardiac imaging, perfusion techniques, high-pitch CT, and dual-energy CT have been introduced to clinical practice, increasing the overall amount of radiation applied.[34] In contrast, a multitude of dose reduction methods have been developed on the hardware side, such as dose modulation along the x-, y-, and z-axes.[5678] Shielding[91011121314] demonstrates a further major step in manufacturer-dependent radiation protection. A recent vast improvement toward ultra-low-dose CT was the implementation of iterative reconstruction (IR) algorithms in the most recent scanner generations. IR has the potential to lower the amount of radiation applied significantly and to maintain the radiation at a constant minimum.[1516] The introduction of the most recent fully integrated circuit detector (ICD) geometry, realized in the Stellar detectors by Siemens (Erlangen, Germany), demonstrated initial promising potential to reduce the radiation further[17] by minimizing electronic noise. To lower the image noise, this new detector assembly combines the electronics and the photodiode into a single unit. Additionally, a major concern in dose reduction has been the preservation of diagnostic image quality, measured by noise, the signal-to-noise ratio (SNR), and the contrast-to-noise ratio (CNR).[181920] The previous data have suggested that diagnostic image quality could be maintained while concurrently lowering the radiation dose, by adopting the recent technologies mentioned above.[2122] We hypothesized that the current implementation of an iterative algorithm and the launch of the ICD geometry would result in a sustainable dose reduction that would preserve the diagnostic image quality. We compared 150 standard chest CTs, from clinical routines that were acquired with the standard Siemens scan parameter presettings, with an intravenous contrast application on a Siemens Sensation scanner with filtered back projection (FBP), the most recent Siemens Definition Flash scanner with IR and the Siemens Edge CT scanner, which was also equipped with IR and ICD. Furthermore, we compared the median radiation doses. The image quality was also assessed during the read-out process by determining SNR and CNR measurements.

MATERIALS AND METHODS

This study was approved by the local ethics committee. Because this study was conducted retrospectively, written informed consent from the patients was waived by the committee. A total of 150 consecutive patients from our clinical practice were enrolled in this retrospective investigation. Fifty patients were scanned using FBP, 50 patients were examined applying iterative image reconstruction (IR), and 50 patients were scanned with ICD using IR, which were identified from the patient set.

The patients were selected by applying the inclusion criteria outlined below:

A standard chest CT, with arterial phase after intravenous contrast media injection, must have been performed on all of the patients, using the same standard preset scan parameters as outlined below. All the scans were required to exhibit diagnostic image quality. Patients with artifacts that resulted in non-diagnostic data sets were excluded from the study population. Attention was paid to foreign bodies (i.e. metal or prostheses); therefore, there were no beam-hardening artifacts in the scan range that could have caused a large increase in the radiation dose. Native scans of the chest or multiphase acquisitions were excluded.

A total of 150 patients (98 men, 52 women) were retrospectively enrolled in this study. After applying the inclusion criteria, 136 patients remained in the study population, which consisted of 91 male and 46 female patients. Fourteen patients (eight men, six women) were excluded. In one case, the contrast medium was injected extravascularly. In 12 cases, additional scans with late enhancements were ordered. One patient received additional expiratory scans for the assessment of air trapping, which resulted in the patient's exclusion from the study population. The case mix in all of the groups consisted of patients who were examined for oncological staging, infectious disease, or trauma. After applying the inclusion criteria, the patients (n) remaining in each subgroup (FBP, IR, IR/ICD) were as follows: n = 48 for FBP, n = 44 for IR, and n = 44 for IR/ICD. The mean patient age in the study group at the time of the examination was 51 years (range 8-94 years).

The scan parameters were identical to the manufacturer's standard presettings for a thorax routine: SOMATOM Sensation 64 (24 × 1.2 mm, pitch 0.8, slice thickness 1.5 mm, FBP); SOMATOM Definition Flash (128 × 0.6 mm, pitch 0.6, slice thickness 1 mm, IR); and SOMATOM Definition Edge (128 × 0.6 mm, pitch 0.6, slice thickness 1 mm, ICD and IR), all by Siemens (Erlangen, Germany). Eighty milliliters of intravenous contrast medium (Iopamiro 300, Bracco Suisse SA, Manno) was injected (injection rate: 3 ml/s). Saline flush of 20 ml was applied. Bolus-triggered scan timing was used on all three scanner types. Aortic threshold for triggering was 100 Hounsfield units (HU). In addition, 5 mm axial slice reconstructions were performed on all three scanners. The standard tube potential (kVp) and tube current (mA) settings for the Siemens routine protocol for the thorax (adapted vendor standard) were maintained constantly on all the scanner generations. An automated tube current modulation (Care Dose^® ; Siemens, Erlangen, Germany), using reference mA (ref. mA) to individually adapt the radiation exposure to the patient size, was implemented on all three of the CT scanners. In addition, Care kV was used on the Flash scanner and the SOMATOM Definition Edge. Tube potential on the Sensation scanner was chosen individually for each patient depending on patient size. The measurements for quality assessment were performed on a soft-tissue kernel reconstruction (B30f, I31f). The standard reconstructions also included a lung kernel reconstruction (B70f, I70f) and coronal and sagittal projections. All the image evaluations were performed on the 5 mm axial slice reconstructions. For the iterative image reconstruction on the Flash and Edge scanners, an iteration level of three was specified in all the cases. An iteration level of three provided a reasonable trade-off between a low radiation dose and image noise.[23] All the patients were scanned in supine position, starting at the superior thoracic inlet and covering the entire lung to the lung bases (including the whole lung parenchyma). The scans were performed with deep forced inspiration. The dose used in conjunction with every CT study was listed by the manufacturer in a dose protocol and was represented by the dose-length product (DLP, mGy*cm) provided by the scanners for a 32-cm-diameter phantom for each scan. In addition, the computed tomography dose index per volume (CTDI Vol.) was provided.

For every patient, the chest volume was calculated retrospectively using the formula for cylindrical objects:

(sagittal diameter/2) × (transverse diameter/2) × π × scan length.

RESULTS

Radiation dose

The radiation dose decreased with every technical advance from FBP to IR and further with IR/ICD [Table 1]. The DLP with FBP for the average chest CT was 308 mGy*cm ± 99.6. In contrast, the DLP for the chest CT examination with an IR algorithm was 196.7 mGy*cm ± 68.8 (FBP vs. IR: P < 0.0001). A further decline in the radiation dose was observed with the combination of IR and ICD: DLP was 167.5 mGy*cm ± 54.5 (IR vs. IR/ICD: P <0.0332) [Figure 1]. Corresponding with the reduction in the DLP, the CTDI Vol. decreased from 7.8 ± 2.5 (FBP) to 5.4 ± 2.0 (IR) and finally to 4.5 ± 1.5 (IR/ICD). The P values for these data were also demonstrated to be statistically significant [Table 1]. In other words, the dose reduction, compared to FBP, was 36.1% with IR and 45.6% with IR/ICD.

Table 1 Dose reduction with IR and IR/ICD compared to FBP. Every step in the reduction is significant

Close

Figure 1

Box plot diagram of the dose distribution with FBP (Sens = sensation), Flash (IR), and Edge (IR/ICD). The dose represented by the DLP (in mGy*cm) declines with every technical advance.

Export to PPT

Image quality

Table 2 summarizes the measured values for the signal and noise in air, aorta, bone, and soft tissue for each of the different CT scanners. The signal in bone was significantly higher using IR combined with ICD than using FBP (P < 0.006). The improvement of the bone signal from FBP to IR alone was not significant. The noise in soft tissue was significantly lower with IR/ICD than with FBP (P < 0.001).

Table 2 Signal and noise calculations for filtered back projection, iterative reconstruction, and IR/ICD detector

SNR values of the aorta, bone, and soft tissue improved significantly switching from FBP on the SOMATOM Sensation scanner to IR/ICD in combination (P values ranging from 0.017 to 0.003) [Table 3]. For the assessment of image quality, Table 3 displays the SNR and CNR calculations for air, bone, aorta, and soft tissue. The total image quality based on average SNR measurements increased (ΔSNR FBP to IR/ICD: P < 0.003). There was no remarkable enhancement of CNR in moving from FBP to IR [Table 3]. CNR in the aorta was 11.6 ± 4.4, 12.8 ± 6.2, and 14.1 ± 6.9 for FBP, IR, and IR/ICD, respectively, representing favorable results for the IR/ICD combination. The CNR was 79.2 ± 43.9, 82.3 ± 50.4, and 61.5 ± 41.9 for FBP, IR, and IR/ICD, respectively [Table 3 and Figure 2]. CNR in relation to DLP and scanner type is displayed in Figure 3. Favorable CNR–DLP relation can be observed for the IR/ICD curve.

Table 3 SNR and CNR measurements for FBP, IR, and IR/ICD

Close

Figure 2

CNR in air and the aorta with the three scanners. The CNR in the aorta is constantly rising. Only the CNR in air exhibits a reduction with IR.

Export to PPT

Close

Figure 3

Scatter plot of the CNR/DLP relationship. Potential trend fitting curves are drawn for every scanner. The fitting curves indicate the mean distribution of CNR-DLP relationship for the separate data sets. Comparing the curves, the IR/ICD combination exhibits favorable CNR maintaining minimal DLP. IR and FBP trend curves show almost similar CNR, but DLP is significantly higher utilizing FBP.

Export to PPT

The mean chest volume of the patients in the three study groups did not differ significantly (P = 0.36-0.67): The mean body volume of the first patient group (SOMATOM Sensation) was 21.33 dm³ ± 7.78 standard deviations (SDs), median = 21.28; the mean body volume of the second group (Flash, IR) was 23.84 dm³ ± 6.66 (median = 23.34); in the third group (Edge, IR, and ICD), the chest volume was 20.09 dm³ ± 5.88 (median = 19.63).

DISCUSSION

Our findings clearly support the assumption that the most recent technical developments, namely IR in combination with an ICD design, could significantly lower the radiation dose burden for patients undergoing standard chest CT examinations. This finding was true even with the standard protocol settings from the manufacturer. The newer technology allowed for a lower tube potential, thus explaining the higher bone signal due to the increased photo effect of radiation at the lower tube potential levels.

In our study, we primarily investigated the feasibility of a dose reduction in chest CT by implementing the most recent software and hardware available at the time. On the software side, IR allows low-dose examinations. The radiation savings with the use of IR have been well-documented in the recent literature.[25] The availability of faster hardware components that reduce computing times has resulted in the standardized implementation of IR, especially with the latest CT scanner generation.[26] The possible dose reduction, in comparison with FBP, is evident and is supported by the present data, retrieved from clinical routines performed on this study population. On the hardware side of CT manufacturing, the minimization of electronic noise has been realized with the ICD by Siemens. Assuming that the noise increases with the distance between the photodiode and the chip, the logical approach to overcome this source of noise is to decrease this distance and to fully integrate the entire electronic component directly into the detector unit. Some initial studies have also demonstrated the possibility of minimizing the radiation dose by implementing IR and ICD in the most recent CT scanner generation by Siemens Healthcare (Erlangen, Germany).[17]

Every step in technical improvement resulted in a significant radiation dose reduction, from FBP to IR and from IR to IR/ICD [Table 1] (P < 0.0001). As our results demonstrate, the possible radiation dose reduction on the software side was approximately 36% in changing from FBP to IR (P < 0.0001). Pontana et al., demonstrated almost the same dose reduction in their study, which compared IR and FBP in chest CT exams.[16] Additionally, using three steps of iteration, Pontana achieved a dose reduction of 35%. Further enhancement in radiation dose reduction obtained by implementing the integrated detector design and changing from the use of FBP to that of an IR/ICD detector was 45.6% (P < 0.0001), indicating that a further reduction of 9.4% from IR to IR/ICD (P < 0.033) was possible with a more effective detector design. These findings are in accordance with the results published by Christe et al., in a phantom-based setting. Implementing iterative reconstruction technique along with the ICD (Stellar detector by Siemens), a dose reduction between 27% and 70% was feasible in this phantom study.[27]

As indicated by the present data, the most evident reduction in dosage was achieved, however, when using IR in contrast to FBP (P < 0.0001). This major step in dose reduction could be explained by the higher dose efficiency of iterative calculations and by the recent methods for accelerating statistical reconstruction.[28]

Not only was the dose reduction between the scanners evident, but also there was an increase in the image quality of the routine clinical CT images as a side effect. At least the radiological parameters of the signal to contrast and noise improved in soft tissues, the aorta, and bone (further studies are required to investigate the other physical parameters of image quality, such as spatial resolution and Z-sensitivity). Nevertheless, regardless of the tissue for which the SNR and CNR were calculated, these parameters increased (aorta, bone, or soft tissue). In fact, there is little use in calculating SNR of air/lung since the more negative these values get, the better it is for the radiologists (ideally, air should present with no signal). Therefore, SNR in the lung is worse for IR/ICD, which could influence the detection of ground-glass opacities to the worse. More importantly, the CNR does not vary significantly between the different scanner types, which are more relevant for disease detection. With SNR and CNR representing image quality, the present data indicated that the overall image quality increased as the dosage concurrently declined. A further dose reduction would therefore be possible if the improvement in image quality was sacrificed for radiation protection. Studies have demonstrated that a loss of image quality due to lower radiation levels did not automatically result in the loss of the diagnostic image quality.[29] Kalra et al., used this effect in their study investigating the sensitivity and specificity of abdominal CT, using IR in ultra-low-dose CT. They were able to conclude that a dose reduction of up to 74% was feasible, while conserving diagnostic image quality, with even less noise than in FBP.[30]

The signal represented in HU increased in bone and soft tissue with IR/ICD compared to FBP. In contrast to this increase, the signals in air and in the aorta declined. Nevertheless, the overall image quality still increased. This phenomenon can be explained by the proportions of signal and noise being significantly more favorable toward increasing SNR values. In contrast to the increased SNR, the CNR exhibited a statistically non-significant increase with the use of IR and the ICD assembly (CNR aorta: P < 0.062 and CNR in air: P < 0.069).

The positive effects of the iterative algorithms can clearly be affirmed, compared to FBP. Regarding the use of the ICD and iteration, one possible gain was the further beneficial radiation minimization, along with enhanced image quality in the low-dose imaging spectrum. The current data set provided a significant result for the improvement of image quality, as emphasized by the SNR values in air, the aorta, and bone tissue.

CONCLUSION

Every step in technical improvements, including iteration and ICD geometry, has resulted in a significant reduction in radiation exposure, with a constant or better image quality level. In conclusion, we can state that the implementation of these recent developments, namely IR and the ICD, indicates possible further dose reduction in chest exams, without a loss of diagnostic image quality.