Patients included in the study were scanned by use of a second-generation 128-slice DS-DECT scanner (Somatom Definition Flash, Siemens Healthcare, Forchheim, Germany). The two x-ray tubes in the scanner were set to operate at 80 and 140 kV (with added tin filtration) with corresponding reference tube current-time products (mAs) of 105 and 41 mAs. These reference mAs levels represent approximately one-fourth of the manufacturer-prescribed reference mAs used for routine DS-DECT. This was done to reduce the radiation dose. The CareDose4D (Siemens Healthcare) was used to further adapt tube current to individual body habitus. Collimation was 64 mm×0.6 mm with z-flying focal spot and the pitch was 0.7.

Weighted average 1-mm thick images (weighting factor, 0.3) were reconstructed by using a standard soft-tissue kernel (Q30f). The dual-energy (DE) ratio and weighted average HU value of each calculus were recorded by manually placing circular regions of interest (ROIs) on the calculi (Fig. 1). The DE ratio was obtained by dividing the attenuation value of the calculus at 80 kV by its attenuation value at 140 kV. Circular ROIs were placed in the aortic lumen at the level of the renal pelvis and in the urinary bladder lumen at the level of the femoral head. The standard deviation (SD) was recorded in these ROIs as a quantitative measure of image noise. In addition, waist circumference was measured on axial sections at the level of the midpoint between the lower margin of the last rib and the top of the iliac crest. Radiation dose information for each study was recorded.

Two radiologists blinded to the patient's details were provided the images to assess the quality of the weighted average images subjectively. The subjective rating of study quality was based on the radiologists' confidence in detecting the calculi, related soft tissue attenuation changes such as perinephric stranding and soft tissue ureteric rim, subjective assessment of the amount of image noise, and image sharpness. By use of these criteria, the two observers rated the studies as being good, average, or poor in quality.

Stones extracted during percutaneous nephrolithotomy or ureteroscopy were subjected to Fourier transform infrared spectroscopy (FTIRS). All stone fragments extracted during the procedure were submitted for analysis. The spectrum was recorded in the SpectrumTwo FTIRS device (Perkin Elmer, Buckinghamshire, UK). According to the FTIRS results, the calculi were classed into five groups: mixed calcium oxalate monohydrate and dihydrate (COM+COD), COM, carbonate apatite (CA), struvite, and UA. The calculi were then retrospectively compared with their HU values and DE ratios.

Analysis of variance (ANOVA) was used to determine if significant differences existed between the DE ratio values of each calculus group. The HU values of each calculus group were subjected to similar analysis. Receiver operating characteristic (ROC) curve analysis was used to identify optimal cutoff DE ratio values to differentiate the various groups of calculi. Image noise and patient radiation dose were compared with patient waist circumference by use of correlation analysis.

The mean calculus size was 15.2 mm (range, 4-31 mm; SD, 5.9 mm). There were 29 calculi measuring <10 mm in diameter, of which 28 were calcium oxalate calculi and one was a CA calculus. All of these calculi were correctly classified as calcium-containing calculi by using DE ratio values. The majority of the stones (~97%) were located in the kidneys; five stones were located in the ureters. All stones in the ureteric location were calcium oxalate stones and were classified correctly by using the DE ratio.

The distribution of the HU values and DE ratios of the calculi is depicted in Table 1. ANOVA analysis showed that HU values could differentiate only UA calculi and calcium-containing calculi confidently (p<0.001). ROC analysis also showed a significant difference between UA calculi and other calculi (area under curve=0.952) with a cutoff of <615 HU having 94.1% sensitivity and 94.2% specificity for identifying UA calculi.

ANOVA analysis showed that there was a significant difference in the DE ratio between the groups (p<0.001). Uric acid calculi had the lowest DE ratio value, followed by progressively higher values in struvite, calcium oxalate, and CA calculi. There was no overlap in the DE ratio of UA and other calculi, with a DE ratio of <1.22 having 100% sensitivity and specificity for identifying UA calculi. The results of the ROC curve analysis of the DE ratios used to differentiate between struvite, calcium oxalate, and calcium carbonate and the further attempted subclassification of calcium oxalate calculi are depicted in Table 2.

The mean waist circumference of the study population was 85.25 cm (SD, 11.2; range, 66-110 cm). The distribution of waist circumference is depicted in Fig. 2. The image quality was assessed by two observers independently and was scored as good, average, or poor. Observer 1 rated 38% of studies as good quality and 62% as average quality. Observer 2 rated 63% of studies as good quality and 37% as average quality. None of the studies were rated as being poor or uninterpretable in quality by either observer. Interobserver agreement was fair (kappa=0.339). The distribution of study quality rating according to waist circumference is depicted in Fig. 3. From this graph we note that high waist circumference did not have a detrimental effect on study quality.

The noise measured in the aortic lumen showed only a weak positive correlation with the waist circumference (Pearson correlation coefficient=0.283, p<0.05). Noise measured in the urinary bladder did not show a significant correlation with waist circumference.

Radiation dose parameters were recorded in 45 patients. There was a significant correlation between the waist circumference and the volumetric computed tomography dose index (CTDIvol) with a Pearson correlation coefficient of 0.917 (p<0.01). The mean CTDIvol was 2.4 mGy (SD, 0.7). The mean dose length product (DLP) was 103.13 mGy·cm (SD, 27.9). The mean effective dose was obtained by multiplying the DLP by a weighting factor of 0.018 mSv/mGy·cm [13], which yielded a mean effective dose of 1.85 mSv (SD, 0.5).

In this study, DECT was accurate in differentiating UA, struvite, and CA calculi from calcium oxalate calculi. Noninvasive therapies like urine alkalinization can be offered to patients with UA calculi [1]. Both struvite and CA stones are associated with urinary tract infections [1415]. For stones associated with infection, the risk of posttreatment sepsis and recurrence of calculi is higher [15]. Hence, identification of these calculi in vivo may help in directing appropriate therapy even before stone extraction.

The discriminant DE ratio values obtained in our study for identifying calculus composition are depicted in Fig. 4. These values are similar to those reported by a phantom study performed by Qu et al. [8] (<1.19 for UA, 1.19-1.46 for cystine, 1.46-1.60 for struvite, 1.60-1.71 calcium oxalate and struvite, and >1.71 for CA and hydroxyapatite) except for struvite stones. This may have been due to the small sample size of struvite calculi (n=3 in our study and n=4 in the phantom study). The study by Acharya et al. [10] is the only previous in vivo study (to the best of our knowledge) performed in the second-generation DS-DECT scanner that has shown the possibility of subdifferentiating non-UA calculi. However, the tube voltage setting used in their study was 100/140 Sn (tin filter) compared with 80/140 Sn used in our study. Whereas Acharya et al. [10] showed that the cutoff DE ratio value of <1.335 had 100% specificity for detection of COM calculi resistant to extracorporeal shock wave lithotripsy, we could not differentiate pure COM from mixed COM+COD calculi in our study.

In our study, we included calculi of mixed composition, with the caveat that the major constituent should contribute to at least two-thirds of the total mass of the calculus. Most studies done earlier included only pure stones or mixed stones in which the major constituent contributed to at least 90% of the total, which does not reflect clinical reality. Thus, our study sample is likely to be more representative of the patient population with urolithiasis.

Radiation exposure in routine CT studies of the abdomen varies between 3 and 25 mSv [16]. Carcinogenesis induced by radiation in the diagnostic energy range is a highly controversial issue. It has been shown that radiation exposure in range of 10 to 100 mSv in atomic bomb survivors is associated with increased risk of carcinogenesis [17]. With epidemiological data lacking for the effects of radiation doses below 10 mSv, the linear no threshold model, which states that there is no safe threshold radiation dose below which radiation-induced carcinogenesis does not occur, has been followed. Relatively few studies have assessed the role of low-dose DECT for predicting urinary calculus composition [1819], and non-UA calculi could not be confidently differentiated in those studies. Ascenti et al. [20] acquired a routine single-energy low-dose CT followed by a targeted DE acquisition of the region containing the calculus. This protocol is difficult to implement in patients with multiple calculi and moreover requires the presence of a radiologist to review the images while the patient is in the scanner. To the best of our knowledge, our study is the first to assess a low-dose DECT protocol in the second-generation DS-DECT scanner.

The radiation doses obtained in our study represent a 3-fold reduction in radiation dose compared to standard-dose DECT (4.5-7 mSv) and a 2- to 3-fold reduction compared to standard-dose single-energy NCCT (4-5 mSv) [21]. Compared to low-dose single-energy NCCT (0.9-1.9 mSv) [21]-the recommended mode for urinary calculus evaluation-there is only a marginal increase in dose.

Patient size did not have any appreciable effect on subjective image quality (Fig. 3) or on the measured image noise. Thus, this low-dose protocol can be relied upon to produce images of at least reasonable quality even in obese patients

The limitations of our study were as follows. The number of struvite and CA calculi encountered in our study was low and hence the DE ratio values of these calculi need further validation by studies with larger sample size. We did not encounter any cysteine, brushite, or hydroxyapatite calculi. Moreover, we could not assess firsthand the impact of calculus composition detection using DECT on patient management. This would require further studies in which the management decisions in patients with urolithiasis are made on the basis of the composition obtained by DECT analysis.