A cross-sectional analytical study was carried out on 100 patients in the emergency department of Vardhman Mahavir Medical College and Safdarjung Hospital in India from November 2018 to April 2020. Institutional Ethics Committee (IEC) approval (IEC Approval number: 2018-78) was obtained.

Inclusion and Exclusion criteria: Patients who were advised NCCT head for any emergency indication such as loss of consciousness, suspected CVA, severe headache, etc., were included. Patients undergoing NCCT head for postoperative check status, and patients with NCCT images showing motion blur and metallic artifacts were excluded from the study.

Fifty cases were selected by systematic random sampling where the first patient on every Monday underwent low-dose CT with reconstruction using both SAFIRE (Group-A) and FBP (Group-B). The subsequent patient presenting on the same day for emergency NCCT head who fulfilled the inclusion criteria and was within five years of age of the previous patient underwent standard dose CT reconstructed with FBP (Group-C). This patient served as the age-matched control of the previous subject.

Data acquisition and reconstruction: Scanning was performed on Siemens 256-slice Dual-source CT scanner. Image acquisition parameters were as follows: slice thickness 0.6 mm, pitch 0.55, FOV 200 mm, collimation 1×1 mm, rotation time 1s. In controls, kVp was set at 120 for adults and 100 for children whereas mAs was set at 390 for adult and 338 for children. In cases, kVp and mAs were automatically calculated by integrated automatic exposure control device Care Dose 4D and Care kV. For standardisation purposes, SAFIRE strength level of three was used in all cases.

Radiation dose measurements: CTDI_vol, DLP and effective dose of all cases and controls were recorded from the dose report available with every scan. The mean effective doses were obtained by multiplying the DLP with tissue factor for head i.e., 0.0021 mSv/(mGy×cm) according to European commissioned published guidelines [7].

Quantitative image analysis: Quantitative image analysis was performed by a single reader blinded to the image protocol. Images were anonymised for Region of Interest (ROI) placements. A 4 mm sq. Region of Interest ROI was placed in GM in the lentiform nucleus, WM in the centrum semiovale and CSF in the lateral ventricle [Table/Fig-1,2 and 3]. Signal was measured as the CT density in Hounsfield Unit (HU) and noise was measured as the Standard Deviation (SD) of attenuation in the ROI. SNR and CNR was then calculated using following formulas:

Mean image noise was calculated as an average of sums of individual SDs of attenuation values in cases and controls, respectively. [Table/Fig-1]:

A 4 mm sq ROI placement in right lentiform nucleus in Grey Matter (GM) in low dose SAFIRE subset [Table/Fig-1a]; and in standard dose FBP subset [Table/Fig-1b].

[Table/Fig-2]:

A 4 mm sq ROI placement in right centrum semiovale in White Matter (WM) in low dose SAFIRE subset [Table/Fig-2a]; and standard dose FBP subset [Table/Fig-2b].

[Table/Fig-3]:

4 mm sq ROI placement in right lateral ventricle in low dose SAFIRE subset [Table/Fig-3a]; and standard dose FBP subset [Table/Fig-3b].

Qualitative image analysis: Qualitative image analysis was performed between the cases (low dose CT reconstructed with SAFIRE) and controls (standard dose CT reconstructed with FBP) on a 4-point Likert scale as detailed in [Table/Fig-4]. Images were analysed by two radiologists with 4 and 20 years of experience and any differences were resolved by consensus.

The data for all patients was entered in MS Excel Spreadsheet and analysed using SPSS version 21.0. Intraobserver variation for the repeated attenuation measurements was assessed with the coefficient of repeatability. Comparison for quantitative image assessment between standard dose-FBP and low dose-SAFIRE images was done by unpaired t-test. Comparison for quantitative image assessment between low dose FBP and low dose SAFIRE images was done by paired t-test. Significance level was taken at p<0.05.

In the present study used two different groups of subjects as cases and controls for ethical reasons. To perform the same group comparison, many of the previous authors have used the data from a single tube of a dual-source CT scanner for reconstructing half-dose images [8-10]. This is possible with a dual-source CT scanner which allows comparison of full-dose and half-dose CT images in the same patient keeping other parameters similar. The downside to this method is that it only allows an absolute reduction of 50% and does not allow any further exploration into dose reduction. To overcome this factor in the present study, we used automatic exposure control called Care Dose 4D and Care kV to reconstruct low-dose images. Care Dose 4D system performs automatic tube current modulation according to patient size and x-ray attenuation changes together with real-time tube current modulation during each tube rotation. CARE kV is designed to automate kV selection to optimise the CNR for a given diagnostic task. This methodology of low-dose CT acquisition appears slightly superior as it avoids unnecessary over or underexposure in subjects with different body habitus.

Majority of the previous authors who did intra-patient dose comparisons like Mitsumori LM et al., and Marin D et al., employed small sample sizes for ethical reasons and limited availability of patients fulfilling inclusion criteria [11,12]. On the other hand, May MS et al., incorporated a larger sample population as they used data from one tube to reconstruct half-dose CT images [8]. The present study employed a large sample size as we utilised vendor-specified pre-set dose reduction protocols and used two different age-matched populations as cases and controls. Most of the previous authors based their study on either adult or pediatric population [11,13-16]. This study included patients from all age groups with age range of subjects varying from 1 to 90 years.

The present study showed that images reconstructed with SAFIRE had better subjective and objective image quality parameters compared to images reconstructed with FBP at both standard dose and low dose with a significant amount of noise reduction. Within the low-dose images, improved SNR and CNR values were seen when reconstruction was done using SAFIRE suggesting that improvement in low-dose CT image quality was due to IR.

Korn A et al., conducted a study on head NCCTs and found an increase of 28% in GM SNR, 31% in WM SNR, and 25 % in the CNR between low-dose CT and standard-dose CT [2]. This study found an increase of 34%, 51.3%, and 43.1% in GM SNR, WM SNR and CNR respectively suggesting that IR not only compensates for the degradation in image quality at lower radiation dose but may even provide a better image quality. Other authors such as Rivers-Bowerman MD and Shankar JJS found higher SNR and CNR values compared to the present study (55% in GM SNR, 35% in WM SNR, and 72% in CNR) in SAFIRE reconstructed scans compared to the FBP reconstructed scans [3].

Majority of the previous studies observed that IR-reconstructed images had a blotchy pixelated appearance which was also seen in current study’s SAFIRE-reconstructed images and this appearance increased with higher strength of iteration from S1 to S5 [15,17,18]. According to Kalra MK et al., lower iterative strengths were associated with a low level of image noise reduction while higher iterative strengths were associated with overly smoothened images [19]. A three-phase phantom-based study performed by Harris MA et al., showed that increasing strengths of IR gave a lower pixel noise and increased CNR, concluding that IR has the potential to affect the pixel noise, image texture, and CNR while maintaining image quality [20]. Due to the above reasons, this study design incorporated a single iterative strength level of 3 for all low-dose CT acquisitions for uniformity. We used this strength level based on the recommendations of the vendor and the results of previous studies. The subjective image quality analysis done in this study showed that the quality of low-dose images reconstructed with SAFIRE was at par with the standard dose images reconstructed with FBP. Based on this we propose that the smoothness of the iteratively reconstructed images can be considered as an inherent property of the type of reconstruction and does not hamper the diagnostic accuracy of the scan.

The present study utilised a second-generation IR method which as per Hwang HJ et al., fared better at improving image quality compared to first-generation reconstruction methods [21]. A previous study by Bodelle B et al., also claims that the conspicuity of ischaemic infarcts was improved with IR [22]. We agree with this observation and also propose that low-dose iteratively reconstructed images are equivalent or even better at demonstrating a myriad of pathological findings such as cerebral infarcts, haemorrhagic contusions, extra-axial bleeds, calcified granulomas etc., which were well demonstrated in the case group.

The present study found a reduction of 39.3% in CTDIvol, a 42.5 % reduction in DLP, and 42% reduction in effective dose in the low-dose subset compared to standard dose subset. The results of dose reduction achieved by various researchers has been summarised in [Table/Fig-10] [2,13,20,23-33].

Korn A et al., studied the potential of IR in NCCT head using SAFIRE. They proposed a 20% radiation dose reduction in NCCT head with mean CTDIvol being 47.8 mGy and mean DLP 829±35 mGy.cm in their low dose dataset [34]. The values in the present study were 29.47±3.53 mGy and 486.49±6.03 mGy.cm, respectively which are significantly less than that obtained by Korn A et al., [34]. This can be explained by the method employed for dose reduction. While Korn A et al., used tube current modulation to obtain low dose dataset, we employed automatic exposure and kV control (Care Dose 4D and Care kV) that amounted to a larger dose reduction. The extent of subjective image quality improvement was also found to be higher in this study [34].

Vorona GA et al., achieved a very low dose reduction of 22% in their study on paediatric head CT. However, their study was conducted on the Alderson-Rando phantom using a different reconstruction algorithm (ASIR) which can explain the variation in results from this study [13].

A few studies such as that by Pickhardt PJ et al., demonstrated aggressive dose reduction to the extent of 74% in abdominal CT [33]. As of today, such aggressive dose reductions are possible only with pure IR methods like MBIR which was used in their study. Hybrid reconstruction techniques like SAFIRE may not be capable of extensive dose reductions but are more suitable for incorporation into routine settings. According to Hara AK et al., a dose reduction of >50% was possible only if image quality in the area outside the ROI was compromised [23].

Scanning time is an important factor while determining the feasibility of routine use of IR technique. The earlier iterative techniques were associated with longer reconstruction times. According to May MS et al., IRIS (first generation IR algorithm) consumed six times the reconstruction time compared to FBP which proved to be a major limitation [8]. However, use of a second-generation IR algorithm SAFIRE did not lead to any significant computational delay in this study. The mean scan time was eight seconds in the thin slice low dose CT reconstructed with SAFIRE and seven seconds in the standard dose CTs reconstructed with FBP suggesting no significant delay in the scan timings between the two methods. This is compatible with incorporation of IR in routine settings. However, it should be emphasised that this is true for only hybrid IR algorithms and not pure IR methods in which the computational time still runs in hours [17,35].

In the present study fared superior to previous studies in a few areas. Firstly, the sample size was larger compared to previously performed studies and we included patients from all age groups from infants to the elderly population. Secondly, a second-generation IR technique that is routinely available in modern CT scanners was evaluated. Hence, incorporation of these in routine settings would not involve additional capital. Thirdly, size-based automatic exposure and kV control which customised the dose reduction to each patient were used.

There were certain limitations of the present study. Due to ethical reasons, the study’s low-dose dataset was obtained on a different set of populations. Hence, the ideal intra-patient comparison was not feasible. The present study did not categorise the extent of dose reduction with respect to the BMI of the individual as dose reduction is expected to be more in people with a lower BMI. The study evaluated dose reduction for only non-contrast CT head whereas the extent of dose reduction varies in different regions of the body as well as in contrast scans.

The IR compensates for image degradation in low-dose CT with improvement in image quality and a radiation dose reduction of 42% is possible in NCCT head without producing any significant delay in scanning, thus allowing incorporation of SAFIRE in routine clinical settings. Even though the study was restricted to only the NCCT head, the results of this study can be extrapolated to other regions of the body as well as to contrast scans. IR has shown dose reduction potential in abdominal scans and CT angiography scans which are associated with high radiation dose. This potential of IR can be applied in CT perfusion studies as well which expose the patients to enormous amount of radiation dose.