Cardiopulmonary Imaging
Original Research
Image Quality of Iodine Maps for Pulmonary Embolism: A Comparison of Subtraction CT and Dual-Energy CT
Audio Available | 
Share
OBJECTIVE. The objective of this study was to compare the image quality of iodine maps derived from subtraction CT and from dual-energy CT (DECT) in patients with suspected pulmonary embolism (PE).
SUBJECTS AND METHODS. In this prospective study conducted between July 2016 and April 2017, consecutive patients with suspected PE underwent unenhanced CT at 100 kV and dual-energy pulmonary CT angiography at 100 and 140 kV on a dual-source scanner. The scanner was set to generate subtraction and DECT iodine maps at similar radiation doses. In 55 patients (30 women, 25 men; mean age ± SD, 63.4 ± 11.9 years old), various subjective image quality criteria including diagnostic acceptability were rated on a 5-point scale by four radiologists and a radiology resident. In 29 patients (17 women, 12 men; mean age, 62.4 ± 11.7 years old) with confirmed perfusion defects, the signal-difference-to-noise ratio (SDNR) between perfusion defects and adjacent normally perfused parenchyma was measured in corresponding ROIs on subtraction and DECT iodine maps. McNemar and Wilcoxon signed-rank tests were used for statistical comparisons.
RESULTS. Diagnostic acceptability was rated excellent or good in a mean of 67% (range, 31–80%) of subtraction CT studies and 36% (5–69%) of DECT studies (p < 0.05 for four of the five radiologists), mainly because of fewer artifacts on subtraction CT. Mean SDNR was marginally higher for subtraction CT than for DECT (18.6 vs 17.1, p = 0.06) and was significantly higher in the upper lobes (21.8 vs 17.9, p < 0.05).
CONCLUSION. Radiologist-judged image quality of pulmonary iodine maps was higher for subtraction CT than for DECT with similar to higher SDNR. Subtraction CT is a software-only solution, so it may be an attractive alternative to DECT for depicting perfusion defects.
Keywords: CT angiography, dual-energy technique, image quality, pulmonary embolism, subtraction technique
Acute pulmonary embolism (PE) is the third most common cause of death from cardiovascular disease [1, 2]. MDCT angiography is the method of choice for PE detection [3]. Various CT techniques that show iodine distribution in the pulmonary parenchyma have been introduced over the last few decades; dual-energy CT (DECT) is the most widely implemented [4, 5]. Several studies have found that iodine maps from DECT accurately depict the effects of acute and chronic PE on parenchymal perfusion, improving detection of small occluding emboli and depicting the extent of pulmonary perfusion defects [6–10]. However, DECT requires a substantial hardware effort.
Subtraction CT is a recently introduced alternative to DECT for generating iodine maps [11]. Subtraction CT subtracts motion-corrected unenhanced images from CT angiography (CTA) studies to obtain maps of iodine distribution in the pulmonary parenchyma [12] (Fig. 1). Because it is purely software-based, subtraction CT can be performed on images from any CT scanner. In addition, subtraction CT is more dose efficient than DECT [13, 14]. However, it is not yet clear whether the image quality of subtraction CT is sufficient for clinical practice, which is the first step in determining whether subtraction CT is capable of showing perfusion defects caused by vascular obstruction with a performance at least equal to that DECT.
 View larger version (339K) | Fig. 1A —64-year-old woman with acute central pulmonary embolism and large right-sided perfusion defect. A, Axial images (3-mm slice thickness) with subtraction CT iodine map overlaid on 100-kV CT angiogram (A) and dual-energy CT iodine map overlaid on mixed 100-and 140-kV CT angiogram (B). |
 View larger version (331K) | Fig. 1B —64-year-old woman with acute central pulmonary embolism and large right-sided perfusion defect. B, Axial images (3-mm slice thickness) with subtraction CT iodine map overlaid on 100-kV CT angiogram (A) and dual-energy CT iodine map overlaid on mixed 100-and 140-kV CT angiogram (B). |
Therefore, in this prospective study, we conducted an intraindividual subjective and objective comparison of image quality of iodine maps derived from subtraction CT and DECT in patients with suspected PE.
| Subjects and Methods |   |
|---|
This prospective study was conducted at a single site, is registered at www.clinicaltrials.gov (NCT02890706), and was approved by the regional ethical review board with written informed consent obtained from all participants (NL56542.091.16). All patients with suspected PE for whom CTA was requested according to the Wells criteria for PE between July 2016 and April 2017 were eligible. Exclusion criteria were age younger than 35 years old (women below this age do not undergo a DECT scan due to radiation dose concerns in standard clinical practice), pregnancy, and hemodynamic instability. Six-month follow-up was used to obtain additional information for the reference standard to determine which patients actually had PE. Figure 2 shows the flowchart of patient inclusion for both image quality analyses.
 View larger version (55K) | Fig. 2 —Study flow chart. For five excluded patients with incorrect reconstructions, data were reconstructed using 3-mm slices instead of 1-mm slices, and raw data had already been removed from CT workstation. Inapplicable dual-energy (DE) iodine map for excluded patient with acute pulmonary embolism (PE) showed major artifacts that rendered map clearly beyond diagnostic acceptability. |
Patients underwent DECT angiography of the chest for clinical evaluation on a dual-source scanner (Definition Flash, Siemens Healthineers), with the addition of an unenhanced CT acquisition for this investigation. Exposure parameters were set to generate iodine maps from subtraction and dual-energy techniques at similar radiation doses. After positioning, an unenhanced scan at 100 kV (reference tube current–time product, 66 mAs; rotation time, 0.50 s; pitch, 0.55) was obtained, followed by IV injection of 60 mL of 350 mg I/mL iodinated contrast medium (Xenetix, Guerbet) and 40 mL of a saline chaser at 5 mL/s. After bolus triggering at the level of the pulmonary artery with a relative threshold of 60 HU, patients were scanned in dual-energy mode (100 and 140 kV; tin filter; reference tube current–time products of 89 and 76 mAs).
All images were reconstructed with a Q30f kernel at 1-mm thickness. DECT iodine maps were generated with the syngo.via workstation Lung Analysis application with default blending (version 3, Siemens Healthineers). Subtraction CT iodine maps were generated using a lung algorithm (SURESubtraction, Canon Medical Systems) that subtracts unenhanced scans from contrast-enhanced 100-kV scans after motion correction [15]. Iodine maps from both techniques were generated with gray-scale and identical color look-up tables, ranging from blue (low iodine enhancement) to yellow (high iodine enhancement).
The scans of the first 60 consecutively included patients were used for the subjective image quality observer study: Five scans were used for observer instruction, and the remaining 55 scans constituted the actual evaluation sample. Four radiologists with 2, 10, 19, and 23 years' experience in chest radiology and one radiology resident served as observers. Among these five readers, two have clinical experience with DECT and three with subtraction CT. Readers were blinded to patient characteristics and technique. Scoring of image quality of the iodine maps was performed on the dedicated scoring workstation. In a single session, observers scored both the subtraction CT and DECT iodine maps of all 55 patients in random order. Using a 5-point scale (1, excellent; 2, good; 3, moderate; 4, poor; 5, extremely poor), readers rated image quality of iodine maps for the following five categories: absence of motion artifacts, absence of other artifacts, parenchymal contrast enhancement, lung segmentation (correctness of the lung mask used for the iodine maps), and diagnostic acceptability.
For the objective image quality evaluation, all scans showing acute PE causing significant vascular obstruction and with presence of perfusion defects were selected. Selection was performed by consensus of an expert panel consisting of three radiologists with access to all images, clinical follow-up, and online patient questionnaires based on the work of Krestan et al. [16] completed 6 months after image acquisition. All scans with ≥ 75% vascular obstruction according to Mastora et al. [17] and a visible correlating perfusion defect without pulmonary infarction on DECT iodine maps were selected. Lungs with emphysema, consolidations, or central PE were excluded.
One radiologist annotated all perfusion defects and the adjacent normally perfused lung areas, preferably in the same lobe, at the same height in the anteroposterior direction on DECT. A researcher placed wedge-shaped ROIs in the annotated regions on CTA displayed with a window width and level setting of 1500 HU and −500 HU, respectively, avoiding large vessels and fissures and transferred them to the DECT and subtraction CT iodine maps. Mean attenuation within the ROIs was calculated excluding zero-valued voxels to avoid locations that had been automatically masked from the iodine maps to exclude small vessels or lesions.
Noise was quantified by subtracting circular ROI attenuation measurements of adjacent slices in homogeneous, normally perfused lung areas and dividing the SD of the result by the square root of 2. Signal-difference-to-noise ratios (SDNR) were calculated by subtracting the attenuation of the perfusion defect ROI from that of the normally perfused lung tissue ROI and dividing the result by the calculated noise.
For each category of subjective image quality, the incidence of each rating was calculated over all cases and over the five observers as a percentage. A cutoff score was used to categorize image quality as acceptable (scores 1 and 2) or unacceptable (scores 3–5). The proportions between subtraction CT and DECT rated as being acceptable were statistically compared with the McNemar test. Mean SDNRs were calculated per lung and per lobe for both techniques. The Wilcoxon signed-rank test was used to test differences between SDNR measurements for the two techniques. Dose-length products (DLPs) were extracted for both techniques and compared with a t test. All statistical analyses were performed using SPSS software (version 22, IBM). Any p value less than 0.05 was considered statistically significant.
| Results | |
|---|
None of the patients who were eligible for the subjective image quality study had incorrectly executed scans or reconstructions. Figure 3 shows the histograms of the mean ratings across all 55 patients (30 women, 25 men; mean age, 63.4 ± 11.9 years old; age range: 39–89 years old) and all observers. For all image quality categories, more cases were rated as excellent or good with subtraction CT than with DECT. The two observers experienced in DECT scored diagnostic acceptability higher for subtraction CT iodine maps than for DECT iodine maps. Two of the three observers experienced in subtraction CT also scored diagnostic acceptability higher with subtraction CT. The last observer scored subtraction CT and DECT similarly.
 View larger version (43K) | Fig. 3A —Histograms of observer ratings for subjective image quality criteria (5-point scale; 1 = excellent, 5 = extremely poor) across all cases and observers. A, Absence of motion artifacts. |
 View larger version (43K) | Fig. 3B —Histograms of observer ratings for subjective image quality criteria (5-point scale; 1 = excellent, 5 = extremely poor) across all cases and observers. B, Absence of other artifacts. |
 View larger version (43K) | Fig. 3C —Histograms of observer ratings for subjective image quality criteria (5-point scale; 1 = excellent, 5 = extremely poor) across all cases and observers. C, Parenchymal contrast enhancement. |
 View larger version (44K) | Fig. 3D —Histograms of observer ratings for subjective image quality criteria (5-point scale; 1 = excellent, 5 = extremely poor) across all cases and observers. D, Lung segmentation (correctness of lung mask used for iodine maps). |
 View larger version (43K) | Fig. 3E —Histograms of observer ratings for subjective image quality criteria (5-point scale; 1 = excellent, 5 = extremely poor) across all cases and observers. E, Diagnostic acceptability. |
Only one error occurred during generation of the subtraction CT iodine maps: an incorrect depiction of the lung apex. This patient was not excluded, and the lung segmentation scored as good by one observer, poor by one observer, and extremely poor by three observers. Table 1 shows the proportion of acceptable image quality ratings for subtraction CT and DECT iodine maps and the number of observers for which one modality was rated significantly higher. When all reader scores were pooled, subtraction CT showed a significantly higher rate of acceptable image quality ratings than did DECT for all categories except contrast enhancement. Figure 4 shows a case in which the DECT iodine map was scored as having more artifacts, and Figure 5 shows a case in which contrast enhancement was rated worse on subtraction CT iodine maps.
 View larger version (345K) | Fig. 4A —78-year-old man with emphysema in upper lobes and no pulmonary embolism. A, Coronal view of subtraction (A) and dual-energy (B) CT iodine maps shows more artifacts in B (arrows). Most of artifacts in B are probably caused by 90° offset between 100- and 140-kV tubes and influx of highly concentrated contrast material in superior vena cava. Diagnostic acceptability of A was scored as excellent by three observers and as good by two. Diagnostic acceptability of B was scored as excellent by one observer, as good by one observer, and as moderate by three observers. |
 View larger version (363K) | Fig. 4B —78-year-old man with emphysema in upper lobes and no pulmonary embolism. B, Coronal view of subtraction (A) and dual-energy (B) CT iodine maps shows more artifacts in B (arrows). Most of artifacts in B are probably caused by 90° offset between 100- and 140-kV tubes and influx of highly concentrated contrast material in superior vena cava. Diagnostic acceptability of A was scored as excellent by three observers and as good by two. Diagnostic acceptability of B was scored as excellent by one observer, as good by one observer, and as moderate by three observers. |
 View larger version (368K) | Fig. 5A —58-year-old man without pulmonary disease, with poor contrast enhancement in pulmonary artery. A, Subtraction (A) and dual-energy (B) CT iodine maps. Image quality ratings for contrast enhancement were lower in subtraction CT iodine map. Diagnostic acceptability of A was scored as poor by three observers and as extremely poor by two observers. Diagnostic acceptability of B was scored as good by one observer, as moderate by one observer, as poor by one observer, and as extremely poor by two observers. |
 View larger version (329K) | Fig. 5B —58-year-old man without pulmonary disease, with poor contrast enhancement in pulmonary artery. B, Subtraction (A) and dual-energy (B) CT iodine maps. Image quality ratings for contrast enhancement were lower in subtraction CT iodine map. Diagnostic acceptability of A was scored as poor by three observers and as extremely poor by two observers. Diagnostic acceptability of B was scored as good by one observer, as moderate by one observer, as poor by one observer, and as extremely poor by two observers. |
In all, 72 regions from 46 lungs of 29 patients (17 women, 12 men; mean age, 62.4 ± 11.7 years old; range, 35–84 years old) were included in the objective image quality assessment. Mean noise across all lungs was equivalent for both techniques: 1.45 HU for subtraction CT and 1.46 HU for DECT (p = 0.99). SDNRs per lung and per lobe (of either side) are shown in Table 2. Only the upper lobes showed a significantly higher SDNR on subtraction CT than on DECT (Fig. 6).
 View larger version (179K) | Fig. 6A —62-year-old woman with acute pulmonary embolism. A, Subtraction (A) and dual-energy (B) CT iodine maps in gray scale (same window width and window levels). Signal-difference-to-noise ratio for A was higher than that in B (15.0 vs 11.3). Difference might have been caused by fewer artifacts being visible in subtraction CT iodine map. Arrow indicates perfusion defect in middle lobe caused by pulmonary embolus. |
 View larger version (169K) | Fig. 6B —62-year-old woman with acute pulmonary embolism. B, Subtraction (A) and dual-energy (B) CT iodine maps in gray scale (same window width and window levels). Signal-difference-to-noise ratio for A was higher than that in B (15.0 vs 11.3). Difference might have been caused by fewer artifacts being visible in subtraction CT iodine map. Arrow indicates perfusion defect in middle lobe caused by pulmonary embolus. |
The median DLP for the 55 patients included in the subjective image quality study was slightly but significant lower for subtraction CT than for DECT: 157 mGy · cm (range, 98–261 mGy · cm) versus 165 mGy · cm (range, 106–273 mGy · cm) (p < 0.05). The median DLP for the cases included in the objective image quality study was also slightly but significant lower for subtraction CT: 154 mGy · cm (range, 100–224 mGy · cm) versus 159 mGy · cm (range, 107–234 mGy ·3 cm) (p < 0.05).
| Discussion | |
|---|
To our knowledge, this is the first intra-patient study comparing subtraction CT and DECT iodine maps in a clinically representative study population of patients with suspected PE. The readers perceived the image quality of subtraction CT to be superior to that of DECT. In addition, quantitative comparison of the depiction of perfusion defects showed overall equivalence between both techniques but a higher SDNR for subtraction CT for perfusion defects located in the upper lobes, even at a slightly lower radiation dose, with annotations for perfusion defects based on the DECT iodine maps.
The significantly higher SDNR for the upper lobes for subtraction CT might be due to the presence of fewer artifacts in that region on subtraction CT than on DECT iodine maps. Felloni et al. [18] also found more artifacts in the upper lobes in a quantification study of DECT iodine maps. Some of the artifacts on dual-energy iodine maps may be explained by the use of a dual-source technique in which the two tubes are offset by roughly a quarter rotation [19]. Consequently, the low- and high-energy data do not match precisely, especially in regions with motion, such as the heart, or with rapid change in contrast enhancement, like the inflow veins. This intrinsic misregistration might explain some of the artifacts seen in our study. Artifacts were less pronounced in two consecutive scans that were registered to each other, as in the case of subtraction CT. However, the clinical significance of the statistically superior SDNR in the upper lobes with subtraction CT remains to be investigated.
Observers were blinded to scanning technique through use of identical color scales. Interestingly, all DECT-experienced readers found subtraction CT iodine maps of acceptable quality more often than the DECT iodine maps. Although using the subtraction CT color scale for both modalities could have influenced the DECT-experienced observers, it would have also affected their perception of the subtraction CT iodine maps.
Our study has some limitations. The DECT iodine maps were generated with a 100- and 140-kV setting. We expect that DECT performs better in terms of noise reduction when tube voltages of 80 and 140 kV or even 80 and 150 kV with a tin filter are used [14]. However, we opted to use the DECT protocol that is used in clinical practice at the institution where the acquisitions were performed, for which a 100-kV setting is used for all patients to accommodate those of larger size [20, 21]. This setting guaranteed a uniform scanning protocol across the study population. A more optimized DECT protocol, such as 80-and 140-kV, can be expected to improve iodine signal and thus SDNR [15]. However, low voltage settings are also likely to improve subtraction CT, with subtraction CT showing better iodine separation than DECT at lower voltage levels [13, 14].
Another potential limitation was the manual placement of the ROIs. CTA was initially used to place the noise ROIs, but the location of some ROIs had to be manually slightly shifted to avoid artifacts that were only evident on the iodine maps or to avoid zero-valued voxels. However, slight shifts would have minimal influence because the wedge-shaped ROI covered a large area and perfusion defects are in the centimeter range [22]. The smaller ROIs to determine noise could have been more affected, especially because smoothing is dependent on vendor settings, but the statistical outcome was the same when the signal difference was not divided by the noise.
| Conclusion | |
|---|
Subtraction CT results in superior subjective image quality for pulmonary iodine maps compared with DECT, although SDNRs between perfusion defects and normal lung regions are similar for both techniques. Given that subtraction CT requires no additional hardware but only registration and subtraction software, this technique holds great potential for tasks currently reserved for DECT.
Based on presentations at the European Congress of Radiology 2017 and 2018 annual meetings, Vienna, Austria.
Supported by Canon Medical Systems, which developed the motion-correction software algorithm in this study. Study data and results were generated and controlled by research personnel at Radboud University Medical Center, with no influence from Canon.
| References | |
|---|
