Reduced dose iterative reconstruction versus standard dose filtered back projection in detection of bladder tumors

The purpose was to assess radiation dose, image quality, and diagnostic performance of reduced-dose scanning with iterative reconstruction (IR) compared with standard-dose with filtered back projection (FBP) with CT urography for detection of bladder tumor. This study was prospectively conducted on 21 patients with bladder masses. All patients were subjected to two scanning protocols: protocol A (standard dose with FBP) and protocol B (additional limited scan to the pelvis at delayed phase with low dose with IR). Based on body weight (< or > 80 kg), each protocol was subdivided into 2 protocols A1 (130 kVp) and A2 (130 kVp) and protocols B1 (80 kVp) and B2 (110 kVp). Radiation dose was assessed in terms of mean CT dose index (CTDI), Dose-length product (DLP) and effective dose (ED). Image quality and diagnostic accuracy were compared in both groups. Mean CTDI, DLP and ED were reduced by average 72.3 % in the 80 kVp protocol (B1) and by 36.3% in 110 kVp (B2) protocol compared to standard-dose protocols. There were significantly lower SNR (signal to noise ratio) between protocol A1 and B1 at aorta and psoas muscles. Subjective image quality analysis revealed no statistically significant differences between the protocol A2 and B2 whereas there were significant differences between protocol A1and B1 as regards to visual image noise and overall image quality. Diagnostic accuracy was identical among different protocols. CT urography with IR scanning showed reduced radiation dose and no difference in detection of urothelial carcinomas from standard dose with FBP despite of degraded image quality in 80 kVp scanning.


Background
CT Urography (CTU) is one-step technique that allows full evaluation of the urinary tract for urothelial tumors, renal masses and stones in a single examination. Meanwhile, in traditional CTU protocols patients receive radiation dose of about 54 mGy, the reduction of radiation dose in CTU has been an important item to be resolved [1,2]. Radiation dose reduction with filtered back projection (FBP) reconstruction techniques is associated with increasing image noises and streak artifacts [3]. With general technical evolution, these drawbacks of FBP have driven the rise of iterative reconstruction (IR) algorithms. The current literature shows that clinical implementation of IR into CT protocols provides essential promise for major improvements in image quality, chiefly noise reduction together with subsequent radiation dose reduction and artifact suppression [3]. The purpose of this study is to compare radiation dose, image quality and diagnostic performance of CTU using reduced-dose scanning with IR versus standard-dose with FBP for detection of urinary bladder tumors.

Methods
Twenty-one consecutive patients presented to Alexandria University Hospitals with sonographically diagnosed urinary bladder masses were enrolled in the study between August 2016 and December 2017.
Enrolled patients had serum glomerular filtration rate greater than 60 ml/min/1.73 m 2 . Patients with contraindications to the use of iodinated contrast media or hip prosthesis were excluded. All patients included in the study were subjected to full history taking including personal data and clinical examination and Routine laboratory investigations.

CT data acquisition
All patients were subjected to two scanning protocols: Protocol A: standard dose with FBP Protocol B: additional limited scan to the pelvis at delayed phase with low dose with IR.

Patients groups
Each protocol is subdivided according to body weight into two protocols: A 1 (130 kVp) and A 2 (130 kVp) and protocols B 1 (80 kVp) and B 2 (110 kVp) based on body weight.
Patients with body weight less than 80 kg (n = 10) to standard dose with FBP 130 kVp (protocol A 1 ) and additional scan limited to urinary bladder at delayed phase with low dose with IR 80 kVp (protocol B 1 ) while patients with more than 80 kg (n = 11) into standard dose 130 kvp (protocol A 2 ) and additional scan limited to urinary bladder at delayed phase with (protocol B 2 ); low dose 110 kvp.
All CT examinations were performed with SOMATOM Perspective 64 MDCT (Siemens Healthcare, Forchheim, Germany) at Alexandria Main University Hospital.
Image acquisition parameters were detector collimation 1.2, slice thickness 1.5 mm, field of view 51 cm, rotation time 0.6 s, helical pitch 0.6, tube current was automatically modulated by the machine, and ranged from 165 to 200 mAs. Additional scan limited to urinary bladder was acquired at delayed phase with scan parameters identical to that of standard dose except for the tube voltage which was reduced to 80 kVp (protocol B 1 ) or 110 kVp (protocol B 2 ) according to body weight.
The injection protocols were the same in all examination, 100 ml of iohexol (Omnipaque 300 mg/ml; GE Healthcare Ireland, Cork, Ireland) which was administered into antecubital vein with flow rate of 2-4 ml/s via a pump injector (Medrad mark V, Bayer healthcare, Pennsylvania, USA). Patients consumed 600 ml water orally while waiting for 1 hour before undergoing CT to allow distention of the urinary bladder. Images were obtained at unenhanced, nephrographic (90-110 s after administration of IV contrast medium), and delayed (420 s after administration of contrast medium) phases for standard FBP protocols, scan range started from above the dome of diaphragm to ischial bones as determined from the AP topogram. Delayed phase only for low-dose protocols limited to the urinary bladder as determined from the prior scans.

Image reconstruction
Raw data of reduced-dose CT examinations in the study group were reconstructed using both mathematical algorithms, FBP and IR (SAFIRE). SAFIRE has 5 preset strength levels. According to recommendations by the vendor, a medium strength level of 3 was applied to all reconstructions in our study. Reconstruction parameters included a B20s + smooth convolution kernel for FBP and the equivalent I30/3+ medium smooth kernel for IR.

Radiation dose report (for the pelvis)
For each patient, volume CT dose index (CTDI) and the dose-length product (DLP) were recorded from the automatically generated patient protocol. The effective radiation dose was then estimated in millisieverts (mSv) (the dose-length product multiplied by the conversion factor of 0.015) [4] for each scan.

Image analysis
Both objective and subjective image quality were assessed for all patients in a blinded and randomized fashion using Syngo via Workplace; Siemens Healthcare.

Objective image quality
The CT attenuation value (in Hounsfield units HU) of aorta and psoas muscle, image noise (standard deviation of CT attenuation values SD), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR) were assessed within the identical two areas (psoas muscle and abdominal aorta) by using a region of interest (ROI) of at least 40 mm 2 at level of L4 before bifurcation of abdominal aorta in delayed phases by one radiologist with 5 years of experience.
SNR and CNR were calculated based on following standard equations, according to Bhosale P., et al. [5]:

Subjective image quality
Subjective image quality and bladder carcinoma were initially assessed independently by two experienced radiologists with more than 20 and 12 years of experience in abdomino-pelvic CT performed visual assessments in blinded randomized fashion. The radiologists rated visual image noise and according to a three-point score (1, unacceptable; 2, average; 3, minimal), and overall image quality according to a 3-point score (1, unacceptable; 2, good; 3, excellent). The presence or absence of bladder lesion was assessed (0, absent; 1, present). Differences between two readers were solved by consensus.
Results were then compared with cystoscopic examination results and pathological analysis.

Demographic patient characteristics
Detailed characteristics of patients were provided in   (Fig. 1).

Objective image quality assessment SNR values
Protocol A 1 vs. Protocol B 1 : There were significant differences in SNRs between protocol A 1 and B 1 at aorta and psoas muscles. Protocol A2 vs. Protocol B2: No significant differences were seen between protocol A 2 and B 2 (P ≤ 0.05 for all). The results of the objective image analyses are listed in Table 3.

CNR values
There was no significant difference at CNR between different groups, as listed in Table 3.

Subjective image quality assessment
Protocol A 1 vs. Protocol B 1 : There were significant differences between protocol A 1 (130kVp) and B 1 (80 kVp) as regards to visual image noise, and overall image quality as listed in (Table 3). Protocol A 2 vs. Protocol B 2 : There were no significant differences between protocol A 2 (130 kVp) and B 2 (110 kVp) as shown in Table 3.

Detection of bladder carcinoma
Diagnosis was identical among different protocols as regards per patient-level analyses among observers. Diagnostic accuracy of CT urography at standard and low-dose protocols was 91.3% (Fig. 2). Detailed results are listed in Table 4. Out of the 21 patients, 19 patients proved to have urothelial bladder cancer (90.4%); 1 of them had urothelial carcinoma with glandular and squamous cell differentiation which is considered bad prognostic factor. One patient had small cell carcinoma (4.7%) and One patient proved to be negative (chronic cystitis) as reported in CTU and proved by cystoscopy and biopsy result. Two cases having multiple superficial small papillary lesions were reported as normal in both standard-and lowdose protocols CT while in cystoscopy, lesions were noted and biopsy results revealed urothelial bladder carcinomas. Per lesion diagnosis was the same for standard-dose and low-dose protocols. Diagnostic accuracy per lesion at standard-dose and low-dose protocols were 78.8%. Superficial small papillary lesions were missed in various cases at CT urography. Detailed results are listed in Table 4.

Discussion
The radiation dose of CT urography is estimated to be approximately 15 to 20 mSv according to the ESUR guideline. The increasing dose awareness in the radiological and the general medical community has driven radiologists as well as equipment manufacturers to optimize CT imaging procedures with the aim of dose reduction [6].
In our study, we compare between standard dose CT with FBP and low-dose CT with IR at delayed phase only. By applying reduced-dose protocols in our study for detection of bladder cancer, CTDI and DLP and ED was reduced by average 72.3 % in the 80 kVp protocol (B 1 ) and by 36.3% in 110 kVp (B 2 ) protocol in comparison with standard-dose protocols, protocol A 1 and protocol A 2 . There is significant decrease in radiation dose, the same as reported in previous studies [7][8][9][10].
In our study, attenuation and noise of tissues of interest were increased significantly when the tube voltage setting was lowered to 80 kVp, which resulted in lower SNR with 3.48 ± 1.56 at aorta and 2.23 ± 0.49 at psoas Fig. 2 Female patient, 63 years old,70 kg. a Axial CT urography with standard dose CT protocol A1, SNR were 4.79 and 3.08 at aorta and psoas muscle respectively and CNR were 9.4 and 8.3 at aorta and psoas respectively. b Axial CT urography with low dose with IR protocol B1, SNR were 2.26 and 2.07 and CNR were 6.6 and 5.6 at aorta and psoas, respectively. Both protocols showing diffuse mural thickening which proved to be TCC  [7,9], SNR of low-dose protocols were higher compared to standarddose protocols. Difference between studies at noise levels and SNR at 80 kVp protocol could be attributed to variability in tube voltage and tube current. While there was no significant difference at CNR between standard-dose protocol (A 1 ) and 80 kVp protocol (B 1 ). As regards to subjective image assessment, increased visual image noise degraded the image quality in 80 kVp, but did not affect the diagnostic accuracy, same as in Bahn et al. [7].
Whereas in 110 kVp, there were no significant differences were noted at SNR and CNR in comparison with standard-dose protocol A 2 . Subjective image analysis was similar to standard dose, for image noise and overall image quality. B 2 protocol showed better lesion conspicuity and even a less image noise and better image quality than standard dose at 18.8% of cases for protocol B 2 as reported by two experienced radiologists. This is consistent with Gervise et al. [9] and Bahn et al. [7] which confirmed that it is possible to perform an acquisition with 100 kV for most patients including overweight patients and obese patients.
Our results showed that the detection of urothelial carcinomas by the reduced-dose images with IR was the same as standard-dose with FBP. Lesion and patientbased accuracies were 78.72% and 91.30%, respectively, among different protocols that had the same range of patient body weight. As reported with Juri et al. [4] and Bahn et al. [7]. The higher accuracies may be attributed to the larger tumor size. Superficial low-grade tumors are still challenging to detect regardless of standard-dose images with FBP or reduced-dose images with IR.

Limitations of the study
This includes small number of patients in this study. A study with larger number of patients is needed. Also, the study only assessed bladder carcinomas, and did not asses other related disease entities, including upper urinary tract abnormalities or lymphadenopathy. Moreover, we did not compare between different CT vendors.

Conclusion
This study showed the use of reduced-dose IR images was associated with significant reduction in radiation dose with no difference at diagnostic detection of bladder carcinomas compared with the use of standard dose FBP images, though IR was associated with an increased noise in 80 kVp images. These results may have important clinical implications using a low-tube voltage scan to reduce cumulative radiation exposure without compromising lesion detection.