Evaluation of split-filter dual-energy CT for characterization of urinary stones

Abstract

Objective:

To assess accuracy of dual-energy computed tomography (DECT) to differentiate uric acid from calcium urinary stones in dual-energy split filter vs sequential-spiral vs dual-source acquisition.

Methods:

Thirty-four urinary stones (volume 89.0 ± 77.4 mm³; 17 calcium stones, 17 uric acid stones) were scanned in a water-filled phantom using a split-filter equipped CT scanner (SOMATOM Definition Edge, Siemens Healthineers, Forchheim, Germany) in split-filter mode at 120 kVp and sequential-spiral mode at 80 and 140 kVp. Additional DE scans were acquired at 80 and 140 kVp (tin filter) with a dual-source CT scanner (SOMATOM Definition FLASH, Siemens Healthineers). Scans were performed with a CTDIvol of 7.3 mGy in all protocols. Urinary stone categorization was based on dual energy ratio (DER) using an automated 3D segmentation. As reference standard, infrared spectroscopy was used to determine urinary stone composition.

Results:

All three DECT techniques significantly differentiated between uric acid and calcium stones by attenuation values and DERs (p < 0.001 for all). Split-filter DECT provided higher DERs for uric acid stones, when compared with dual-source and sequential-spiral DECT, and lower DERs for calcified stones when compared with dual-source DECT (p < 0.001 for both), leading to a decreased accuracy for material differentiation.

Conclusion:

Split-filter DECT, sequential-spiral DECT and dual-source DECT all allow for the acquisition of DER to classify urinary stones.

Advances in knowledge:

Split-filter DECT enables the differentiation between uric acid and calcium stones despite decreased spectral separation when compared with dual-source and dual-spiral DECT.

Introduction

Urolithiasis is a common disease among adults with an increasing prevalence in the recent years.^1–3 Low-dose, non-contrast computed tomography (CT) is the recommended diagnostic tool to examine patients with acute flank pain and suspected urolithiasis. Besides providing a high sensitivity and specificity while being non-invasive, CT also allows to determine the location, size, and attenuation of urinary stones.^4–6 Furthermore, implementation of the dual energy CT (DECT) enables material composition analysis based on chemical properties of the urinary stones.^5,7,8 Material composition analysis with DECT scanning is based on the material attenuation profile at varying tube potential. The dual-energy ratio (DER) represents the attenuation profile by combining attenuation values from low- and high-energy datasets.^5,9

Clinically, information about the stone composition is important for treatment planning. Uric acid stones can be dissolved through alkalization of the urine by oral chemolysis. In contrast to that, calcium stones may require invasive treatments, such as extracorporeal shockwave lithotripsy, ureteroscopy, or percutaneous nephrolithotomy.^10–12

Different dual energy techniques are currently available including dual-source DECT, fast kV-switching DECT, dual-layer DECT, and sequential-spiral DECT. Dual-source DECT was the first dual energy technique that was utilized for material analysis.^4,7,13 Previous studies demonstrated the capability of fast kV-switching^14,15 and dual-layer¹⁶ DECT to differentiate urinary stone materials. Recently, Nakhostin et al. compared different split-filter generations and dual-source DECT regarding their accuracy of characterizing different urinary stone materials¹⁷. The aim of our study was to investigate the ability of split-filter DECT to differentiate between calcium and uric acid urinary stones.

Methods and materials

Phantom setup

Thirty-four pure human urinary stones (17 calcium, 17 uric acid) were retrieved through either spontaneous passage, ureterorenoscopy, or percutaneous nephrolithotomy. Their composition was assessed through infrared spectroscopy. All stones were hydrated in saline for 24 h and, together with saline, placed in a plastic vial. The vials were placed in the center of a water-filled cuboid polyethylene phantom (30 × 22 × 30 cm³) by distributing them along the x and z (horizontal) axes of the CT scanner. The phantom was placed in the center of the CT gantry for scanning.

CT protocol

The stones were scanned in three different scanning modes on two different scanners. Split-filter dual-energy CT is a single source dual-energy technique provided by the CT scanner SOMATOM Definition Edge (Siemens Healthineers, Forchheim, Germany). It uses a gold and a tin filter to separate the 120 kVp X-ray beam into high (tin) and low (gold) energy photon spectra.¹⁸ The same scanner also has the capability to perform sequential-spiral DECT (“dual-spiral mode”) in which two successive CT scans at different energy levels are performed, in this case at 140 kVp and 80 kVp (without additional filtration).

The third scan served as reference standard and was obtained with a dual-source DECT scanner (SOMATOM Definition FLASH, Siemens Healthineers, Forchheim, Germany) with two perpendicular mounted tube-detector units that simultaneously acquire images at two different energy levels. For this study, we utilized 80 kVp and 140 kVp. Tin filtration was used to further alter the high energy spectrum.

Both CT scanners feature 128 rows and a z-flying focal spot. Scans were performed with 64 × 0.6 mm collimation, a pitch of 0.6 and 0.5 s rotation time. Images were reconstructed with a slice thickness of 0.75 mm, an increment of 0.5 mm, and a 353 mm field of view with a 512 × 512 pixel matrix resulting in a pixel size of 0.69 mm. The CTDIvol across all protocols was kept comparable at 7.3 mGy by manually adjusting the tube current-time product.

Analysis of urinary stones

Using the MATLAB environment (R2014a, Mathworks, Natick, MA), a dedicated software for stone analysis was created by one of the co-authors. The script enabled a 3D stone segmentation and the measurement of the mean attenuation. The software automatically segmented the stone in the high energy dataset, which was then mirrored to the low energy dataset for a pixel-by-pixel analysis. Average attenuation values at low and high tube potential for every stone were used to calculate the DER using the formula listed below. Size and volume measurements were also performed.

Urinary stone characterization

Urinary stone characterization with DECT is based on the attenuation profiles of materials at different tube potentials. The ratio of attenuation values at high vs low tube potential was used to classify the stones (dual energy ratio, DER = density_{low kVp} / density_{high kVp}).^7,19,20 The material composition was assessed with infrared spectroscopy as reference standard. Stones containing at least 90% of one single material were considered pure.

Statistics

DERs were plotted for each of the scan techniques. The level of statistical significance was set to 0.05. Normal distribution was assessed using the Shapiro–Wilk test. Wilcoxon test was used to compare DERs and attenuation values between calcium and uric acid stones. Spearman coefficient (r_Sp) was used for the assessment of a correlation between stone size and DER values.

Results

Volume and attenuation of urinary stones

The mean volume of urinary stones was 89.0 ± 77.4 mm³ (range 4.0–321.0 mm³). The 17 calcium stones had a mean volume of 115.4 ± 72.7 mm³ and the 17 uric acid stones had a mean volume of 69.8 ± 76.4 mm³ (p = 0.012). Calcium stones had a significantly higher attenuation than uric acid stones (p < 0.001, Table 1).

Table 1.

Urinary stone attenuation of calcium and uric acid stones at high and low energy settings for each dual-energy technique

		calcium stones	uric acid stones	p value
dual-source	80kVp	1344.9 ± 212.1 HU	374.7 ± 103.3 HU	<0.001
	140kVp	814.6 ± 131.9 HU	404.1 ± 110.1 HU	<0.001
sequential-spiral	80kVp	1124.2 ± 215.0 HU	295.9 ± 100.3 HU	<0.001
	140kVp	805.7 ± 134.6 HU	333.0 ± 99.2 HU	<0.001
split-filter	120kVp + tin	914.4 ± 144.9 HU	363.7 ± 93.5 HU	<0.001
	120kVp + gold	756.6 ± 118.8 HU	355.9 ± 99.5 HU	<0.001

Characterization of urinary stones

Using the DER, all three DECT modes enabled a correct categorization of uric acid and calcium stones in 34/34 cases (100%). All three DECT techniques led to significantly different attenuation values as well as DERs between uric acid and calcium stones (p < 0.001 for all, Tables 1 and 2Tables 1 and 2, Figure 1). The DERs of calcium vs uric acid stones were 1.20 ± 0.02 vs. 1.03 ± 0.08 (range 1.14–1.27 vs 0.97–1.31) for split-filter DECT; 1.38 ± 0.06 vs. 0.88 ± 0.07 (1.27–1.52 vs 0.71–0.98) for sequential-spiral DECT and 1.64 ± 0.07 vs. 0.93 ± 0.06 (1.54–1.80 vs 0.84–1.09) for dual-source DECT (Table 1). DER values for calcium stones were significantly higher for dual-source DECT compared with split-filter (p < 0.001) and sequential-spiral DECT (p = 0.014).

Table 2.

Dual energy ratios for calcium and uric acid acquired with different dual-energy techniques

	calcium stones	uric acid stones	p value
dual-source	1.64 ± 0.08	0.93 ± 0.06	<0.001
sequential-spiral	1.39 ± 0.07	0.88 ± 0.07	<0.001
split-filter	1.20 ± 0.03	1.03 ± 0.08	<0.001

Figure 1.

Attenuation at high and low energy of calcium and uric acid stones for dual-source dual energy computed tomography (DECT) (A), sequential-spiral DECT (B) and split-filter DECT (C). Dual-source DECT achieves the best differentiation between the materials whereas the differences are smaller in split-filter dual-energy CT. Cal - Calcium, HU - Hounsfield Units, UA - uric acid

DER of uric acid stones were significantly higher for split-filter DECT (p < 0.001 for sequential-spiral; p = 0.002 for dual-source) while values were comparable between dual-source and sequential-spiral DECT (p = 0.215) (Figure 2).

Figure 2.

Dual-energy ratios of calcium and uric acid stones for all three dual energy scan modes. Spectral separation is best for dual-source DECT.

Split-filter DECT correctly categorized calcium and uric acid stones with a sensitivity, a specificity, positive and negative predictive values of 100%. Compared to dual-energy und sequential-spiral DECT; however, DER boxplots show a reduced spectral separation in split-filter DECT (Figure 2).

Discussion

In this study, we investigated the capability of split-filter, sequential-spiral and dual-source DECT in characterizing urinary stones. Using the dual energy ratio (DER), split-filter, sequential-spiral, and dual-source DECT techniques were able to correctly classify all urinary stones according to their composition.

Different studies have evaluated the value of split-filter DECT. Jacobsen et al. found the lowest accuracy for split-filter DECT when comparing iodine quantification and monochromatic attenuation from different DECT techniques.²¹ May et al. have described a lower spectral separation when comparing split-filter DECT and dual-source DECT in head and neck CT.²² Almeida et al also depict a higher spectral overlap when compared to dual-source DECT.¹⁸ In contrast to Nakhostin et al. the DER values from uric acid and calcified stones in our study did not show any overlap in either acquisition mode.¹⁷ In their study, they include struvite stones into the group of calcified stones. Struvite stones, however, have been reported to have a lower attenuation than calcified stones.²³ Furthermore, we analyzed pure stones consisting >90% of one single material. Nakhostin et al. do not specify the composition of the stones used in their study.

Our results for the capability of dual-source, sequential-spiral and split-filter energy DECT to classify urinary stones are in agreement with several previous studies^{4,5,7,9,13,17,24–28} that have shown nearly 100% specificity and sensitivity in categorizing urinary stones. Most studies have been performed using Siemens Somatom Definition or Definition Flash with or without additional tin filtration.

Additional factors, for example, the definition of stone purity, also compromise the comparability of results. However, Boll et al. assessed stones with at least 95% of one material with 80 kVp and 140 kVp tube potential and found that one mixed uric acid stone could not be differentiated between uric acid and cystine.⁷

Stolzmann et al. utilized the additional tin filter, which improved differentiability compared to scans without tin filter.²⁵ By scanning 110 urinary stones at 80 and 140 kVp and using a semiautomatic software, sensitivity and specificity of 100% were achieved. Another study by Spek et al. also reports one falsely classified uric acid stone as calcified.²⁴ In this case, further chemical analysis resulted in a material composition of 90% uric acid and 10% calcium hydroxyapatite.

Notwithstanding that compared papers slightly differ in image acquisition parameters, stone density values from all DECT techniques were in agreement to the literature when limiting the papers to Siemens Somatom Definition and Definition Flash and 80 or 140kVp.^7,19,24,25

Our study has a number of limitations. First, this was a phantom study to confirm feasibility of the technique for material separation specifically for urinary stones. However, the phantom may not fully reflect the actual patient scan, due to impact of motion,²⁹ beam hardening, body habitus, and other patient factors. Therefore, confirmation of our findings in future in vivo studies is necessary. Second, we did not evaluate the performance of split-filter DECT in categorization of urinary stones of mixed composition. However, the aim of this study was to compare the overall ability of dual-energy scan techniques to differentiate between different types of urinary stones compositions, which reflects clinical routine.

Lastly, an assessment of the overall image quality was not performed. This may be important when evaluating differential diagnoses or incidental findings and, therefore, should be also assessed in the future studies.

In conclusion, split-filter DECT, dual-source DECT, and sequential-spiral DECT enable the characterization of urinary stone materials through dual-energy ratio (DER) in a controlled environment of the phantom. Spectral separation of split-filter DECT appears to be somewhat limited when compared to dual-source or sequential-spiral DECT. Impact and feasibility of this technique should be explored in vivo as a next step.