Size-Specific Dose Estimates Based on Water-Equivalent Diameter and Effective Diameter in Computed Tomography Coronary Angiography.

BACKGROUND To determine the difference in size-specific dose estimates (SSDEs), separately based on effective diameter (deff) and water equivalent diameter (dw) of the central slice of the scan range in computed tomography coronary angiography (CTCA). MATERIAL AND METHODS There were 134 patients who underwent CTCA examination, were electronically retrieved. SSDEs (SSDEdeff and SSDEdw) were calculated using 2 approaches: deff and dw. The median SSDEs and mean absolute relative difference of SSDEs were calculated. Linear regression model was used to assess the absolute relative difference of SSDEs based on the ratio of deff to dw. RESULTS The median values of SSDEdeff and SSDEdw were 18.26 mGy and 20.56 mGy, respectively (P<0.01). The former was about 10.08% smaller than the latter. The mean absolute relative difference of SSDEs was 10.48%, ranging from 0.33% to 24.16%. A considerably positive correlation was found between the absolute relative difference of SSDEs and the ratio of deff to dw (R²=0.9561, r=0.979, P<0.01). CONCLUSIONS The value of SSDEdeff was smaller by an average of about 10.08% than SSDEdw in CTCA, and the absolute relative difference increased linearly with the ratio of effective diameter to water equivalent diameter.

Background

Computed tomography (CT) volume index (CTDIvol) and dose length product (DLP) are widely used in clinical practice to quantify radiation dose from CT scan and they help in performing quality assurance procedures [1-4]. CTDIvol measured in mGy is routinely estimated by using standard 16 cm or 32 cm diameter polymethyl methacrylate cylinder phantoms and is susceptible to scan parameters, such as kV, mAs, pitch, collimator, bowtie filter, and so on. DLP measured in mGy·cm is the product of CTDIvol multiplied by the scan range, and it is the metric of total radiation dose output from a given CT scan. Presently, CTDIvol and DLP are displayed on CT units for each scan [5]. Although these metrics are tagged to individual examination, they do not take into account the correlated factors of patients undergoing CT examination [5,6-8]. Therefore, these 2 metrics represent the radiation dose output of CT scanner with the given scan details, but not the radiation dose absorbed by the patient [5,9-11].

On the basis of a large number of studies on CTDIvol normalized to patient’s geometric size and different attenuations of various substances, the American Society of Physicists in Medicine (AAPM) Report 204 and 220 introduced the concept of size-specific dose estimate (SSDE), which is the product of CTDIvol and size-dependent conversion factor (f) [12,13]. The SSDE corrects the phantom-derived scanner-indicated CTDIvol according to the patient size and more accurately and reasonably estimate the radiation dose at the center of the scan range [10,11].

SSDE metrics were classified as SSDEdeff based on effective diameter (deff) and SSDEdw based on water equivalent diameter (dw). A recent series of articles reported radiation dose to investigate the differences between SSDEdeff and SSDEdw in CT examinations of the torso, such as chest, abdomen, and pelvis [14-16]. These studies demonstrated that SSDEdeff underestimated radiation dose in chest compared to SSDEdw, on the contrary, SSDEdeff was generally greater than SSDEdw in abdomen and pelvis. Due to the different anatomic section, scan range and required contrast medium in CT coronary angiography (CTCA), the discrepancy of SSDEdeff and SSDEdw in the aforementioned studies may not account for that of CTCA. Furthermore, to the best of our knowledge, no report on the 2 SSDE metrics in CTCA has been published so far. The purpose of this work was to assess and compare individual radiation dose metrics of SSDEdeff and SSDEdw at the mid-point of the scan range from patients who underwent CTCA.

Material and Methods

Patient population

This retrospective study was approved by the institutional Ethics Committee and written informed consent was waived. Initially, 162 patients who underwent CTCA examination were electronically queried in Picture Archive and Communication System (PACS) of one institution, Zhejiang Provincial People’s Hospital. Patients who had known allergic reaction to iodine contrast medium, severe renal failure, suspected and known pregnancy were excluded. All patients had clinically indicated or diagnosed coronary artery disease (CAD). There were 28 patients excluded because they had stent implant, mechanical valve replacement surgery, metal bodies on the skin, and truncated images which may result in potential inaccuracy of radiation dose exposed to patients (in SSDEdw). Finally, for the period between January 2018 and June 2018, a total of 134 patients were enrolled in this retrospective study. There were 91 males and 43 females, their mean age was 59.67±11.70 years (range 30 to 90 years), their mean weight was 64.72±9.54 kg (range 44 kg to 90 kg) and their body mass index (BMI) was 23.79±2.57 kg/m2 (range 17.14 kg/m2 to 29.90 kg/m2).

Data acquisition

All patients with a heart rate (HR) <65 beats per minute underwent axial volume CT scan on 320-detector CT (Aquilion ONE, Toshiba Medical Systems, Otaware, Japan). All the examinations were performed within 1 beat acquisition with prospective electrocardiogram-gating. A breath-hold exercise was performed before diagnostic scan. The diagnostic exposure phase window was limited automatically to 70% to 80% of the R-R interval by the scanner on the basis of HR during a breath-hold exercise. The scan parameters were tube voltage 100 kVp to 120 kVp, tube current 400 mA to 550 mA, and rotation time 0.35 seconds per rotation. Tube voltage and current were manually adjusted by radiographer according to individual BMI and shape of the imaging region. The other key parameter was that the scan range matched the personal length along the z axis of the heart, corresponding to four options of 120 mm, 128 mm, 140 mm, and 160 mm. The images were reconstructed with soft tissue algorithm (FC43 kernel), a 512×512 matrix, 400×400 mm FOV, 5 mm of slice thickness, and 5 mm of slice interval. The reconstructed images were automatically transferred to PACS (Greenlander version 6.0, Mindray Healthcare, Shenzhen, China).

A manual trigger technique was used across all patients. A 30 mL saline solution was injected via an 18-gauge catheter placed in the antecubital vein at a rate of 6.0 mL/second to test the injection pressure. This facilitated the decrease of the risk of extravagated contrast medium during contrast medium administration. A dose of 0.6 mL/kg contrast medium with an iodine concentration of 320 mg/mL was injected over 10 seconds using a dual power injection system. Injection of this iodine solution was followed by 20-mL diluted contrast medium with a ratio of 3 to 7 (contrast medium to saline solution) and 30 mL of flush saline solution at the same rate as the contrast medium.

Calculation of SSDE

Deff, as defined in AAPM 204, was the diameter of the maximal anteroposterior and later dimensions. Patient sizes of AP and LAT were manually measured on the central transverse image of the CTCA scan range. AP and LAT values were summed to obtain a single index [12], as follows:

A semi-automated segmentation technique based on CT value threshold, filling holes, keeping largest and editing mask with Mimics software (version 17, Materialise Medical System, Belgium) was used to delineate the boundary of the transverse image out from the surrounding air and the table. The average CT value and total pixel numbers were automatically reported in the properties of the established mask to calculate dw of the axial image. In addition, lower attenuation region (Arealow) with CT value smaller than −600 HU and higher attenuation region (Areahigh) with CT value greater than 200 HU, were segmented out from the transverse image, and the CT values of Signallow and Signalhigh corresponding to Arealow and Areahigh were documented. The value of dw was calculated as [13];

Where CTROI and AROI are the average CT value and the area of the axial image at the central slice location of the scan range respectively. AROI was calculated as;

Where N is the sum of pixels on the axial image while Apixel is the area per pixel in cm2.

SSDEs (SSDEdeff and SSDEdw, respectively) were derived from both deff and dw. SSDE was calculated as;

Where f is the size-dependent conversion factor to correct patient size in deff and dw, and are defined as fdeff and fdw, respectively. CTDIvol reported by the scanner is the average CTDIvol across all slices of the scan range. Due to the tube voltage of 120 kV, f was calculated as [12];

Where d is the value of deff or dw to express patient size in centimeter.

In this work, each scan protocol was conducted using standard 32 cm diameter polymethyl methacrylate cylinder phantoms to obtain the CTDIvol across all slices. According to the special approach for calculating SSDE, the data set was divided into group A and group B. In group A, 134 patients were included, and patient size was characterized by deff. SSDE was defined as SSDEdeff, which was calculated using fdeff at the central slice multiplied by CTDIvol value displayed on the radiation dose page. Similarly, in group B, 134 patients were included, and patient size was characterized by dw. SSDE was defined as SSDEdw, which was calculated using fdw at the central slice multiplied by CTDIvol value. To observe the homogeneity of the body phantom of 32 cm and actual body size, the difference of 32 cm and deff, 32 cm and dw , (32-cm versus deff, 32-cm versus dw) was calculated. The difference was defined as . The absolute relative difference, Erssde, between SSDEdeff and SSDEdw was calculated to observe the accuracy of estimation dose. To study the change of Erssde with dw, patients were split into 4 segments according to interquartile range of water equivalent across all patients. The 4 segments of patients were, dw-segment 1 for dw ≤23.82 cm, dw-segment 2 for 23.82 cm 26.31 cm.

Statistical analysis

All data were tested using Shapiro-Wilk test and Levene test. Numerical data with a normal distribution was reported as mean±standard deviation. Those with a skewed distribution were reported as median (P, P). Student’s 2-tailed t-test was used to compare , body size, area, and signal, while Wilcoxon was performed for f, and SSDE. A broken line graph was used to illustrate the trend of Erssde changing with dw. The difference of CTDIvol, SSDEdeff, and SSDEdw was observed using Friedman test.

Pearson correlation test was performed for SSDEdw and dw, as well as for Erssde and dw, while Spearman rank correlation test was carried out for SSDEdeff and dw, as well as for Erssde and Prosize. Linear regression models were used to estimate the separate relationship of deff and dw, SSDEdeff and SSDEdw, Erssde and the ratio of deff to dw (named as Prosize). Multiple stepwise regression analysis was performed to observe the effect of Arealow, Areahigh, Signallow, and Signalhigh (independent variables) on SSDEdeff and Erssde (dependent variables), respectively. To assess the magnitude of variation explained by independent variable, the squared coefficients of determination (R2) was calculated. A P-value of less than 0.5 was considered to indicate statistically significant difference. All statistical analyses were conducted using statistical software PASW 18.0 (IBM Corp. Armonk, NY, USA).

Results

A total of 134 axial images were measured in this work. There were 133 slices with deff smaller than 32 cm of body phantom, while 1 slice was higher than 32 cm of body phantom. All dw values were smaller than 32 cm. There was no slice with body size equal to 32 cm. All values of fdeff and fdw were greater than 1. There was no slice with f less than or equal to 1.

As shown in Table 1, there was significant difference in , body size and SSDE of the 2 groups. The average deff was about 9.99% higher in group A than dw in group B. The average SSDEdeff was about 10.08% smaller than SSDEdw.

Table 1

Mean and standard deviation of the and Body Size (cm), Median (P25, P75) of f and SSDE (mGy).

Approach	Δd32  (cm)	Body size (cm)	f	SSDE (mGy)
A	4.48±1.75	27.52±1.75	1.33 (1.26, 1.41)	18.26 (15.65, 21.72)
B	6.96±1.80	25.04±1.80	1.48 (1.40, 1.56)	20.56 (17.21, 24.00)
P	0.000*	0.000*	0.000**	0.000**

Approach A – size-specific dose estimate based on effective diameter; Approach B – size-specific dose estimate based on water equivalent diameter. difference of phantom diameter and body size; f size-dependent conversion factor; SSDE size-specific dose estimate;

Student’s t-test;

Wilcoxon test.

The median (P, P) of CTDIvol, SSDEdeff and SSDEdw were 13.15 (interquartile range 11.48, 16.60) mGy, 18.26 (interquartile range 15.65, 21.72) mGy, and 20.56 (interquartile range 17.21, 24.00) mGy, respectively. CTDIvol was about 24.36% (range 8.15% to 39.69%) smaller than SSDEdeff, and about 32.09% (range 24.72% to 47.48%) smaller than SSDEdw. SSDEdeff was about 10.08% (range −2.89% to 24.19%) smaller than SSDEdw. A significant difference was found in these 3 radiation metrics (χ2=264.060, P<0.01). A representative case is shown in Figure 1.

Figure 1

A 64-year-old male undergoing computed tomography coronary angiography (CTCA). His body mass index (BMI) was 22.49 kg/m2. An image at the central location in the scan range. The anteroposterior (AP) and lateral (LAT) were 22.9 cm and 32.9 cm, respectively. The effective diameter (deff) was 27.5 cm. The size-dependent conversion factor (fdeff) was 1.33. The CT volume index (CTDIvol) of this scan was 10.7 mGy. The size-specific dose estimate (SSDEdeff) was 14.26 mGy based on formula 4. The water equivalent diameter (dw) was 24.4 cm and the size-dependent conversion factor (fdw) was 1.52. The size-specific dose estimate (SSDEdw) was 16.30 mGy while SSDEdeff was about 12.52% smaller than SSDEdw.

As shown in Figure 2, deff was positively correlated with dw (R2=0.6434, r=0.802, P<0.01), while SSDEdeff was positively correlated with SSDEdw (R2=0.9436, r=0.972, P<0.01). Arealow and Areahigh were 170.28±45.35 cm2 (range 68.59 to 326.75 cm2), 74.16±11.64 cm2 (range 45.06 to 100.78 cm2), respectively and a significant difference was found between them (t=24.126, P<0.01). Signallow and Signalhigh were −889.56±75.58 HU (range −621.36 to −998.36 HU) and 407.19±37.32 HU (range 326.81 to 527.53 HU), respectively, and there was a significant difference between them (t=−170.699, P<0.01) as well. Multi stepwise regression analysis showed that Signalhigh (normalized β=−0.528) was independently and negatively associated with SSDEdeff. Arealow, Signallow and Areahigh were not included in the regression equation.

Figure 2

(A) Scatter plot representing relationship between deff and dw. There was a considerably positive correlation (Pearson analysis, R2=0.6434, r=0.802, P<0.01). (B) Scatter plot representing relationship between SSDEdeff and SSDEdw. There was a considerably positive correlation (Spearman analysis, R2=0.9436, r=0.972, P<0.01). dw – water equivalent diameter; deff – effective diameter; SSDEdw – size-specific dose estimate based on water equivalent diameter; SSDEdeff – size-specific dose estimate based on effective diameter.

There was a weak positive correlation between SSDEdeff and dw (r=0.267, P=0.002), the same correlation level was found between SSDEdw and dw, however, it was not statistically significant (r=0.136, P=0.116).

The average of Erssde was 10.48±4.76%, ranging from 0.33% to 24.16%. There was a moderate negative correlation between Erssde and dw (r=−0.342, P<0.01). As shown in Figure 3, Erssde changed with dw. Between dw-segment 1 and dw-segment 4, Erssde declined from 11.52% down to 8.22%. There was a considerable positive correlation between Erssde and Prosize (R2=0.9561, r=0.979, P<0.01). With Prosize as a dependent variable, Arealow, Areahigh, Signallow and Signalhigh as independent variables, multiple stepwise regression analysis showed that Arealow was independently and positively associated with Prosize (normalized β=0.504, P<0.01), whereas Signallow was independently and negatively associated with Prosize (normalized β=−0.461, P<0.01). Both Areahigh and Signalhigh were not included in the regression equation and had an insignificant influence on Erssde.

Figure 3

(A) Line chart representing the relationship between Erssde and dw segment. A decreasing trend of Erssde was illustrated with dw increasing. Erssde declined from 11.52% down to 8.22%. (B) Scatter plot representing relationship between Prosize and Erssde. There was a considerable positive correlation (R2=0.9561, r=0.979, P<0.01). Erssde, absolute relative difference of size-specific dose estimates based on effective diameter and water equivalent diameter; dw-segment, patients split according to interquartile range of water equivalent across all patients; dw-segment 1, patients with dw ≤23.82 cm; dw-segment 2, patients with 23.82 cm 26.31 cm. Prosize – the ratio of effective diameter to water equivalent diameter.

Discussion

Compared with SSDE, CTDIvol tends to underestimate radiation dose ranging from 14.29% to 36.46% in CT chest sans, especially for thin or pediatric patients [1,17]. Consistent with previous studies [1,17], the current findings revealed that CTDIvol in CTCA estimated patient dose to be smaller than 27.95% and 37.20% on average than SSDEdeff and SSDEdw, respectively. CTDIvol in torso is obtained on the basis of the standard phantom of 32 cm diameter. In contrast, the actual values of deff and dw in adult chest were almost smaller than 32 cm and f was greater than 1. The difference between 32 cm and actual chest size might result in CTDIvol estimation inaccuracy. Thus, the standard 32 cm diameter polymethyl methacrylate cylinder phantom used to represent realistic adult chest to estimate radiation dose in chest CT examination is controversial [5,15,18,19]. So, instead of CTDIvol, SSDE, which takes patient correlated factors into account, can be considered as a great positive step in the field of CT dose estimation.

Compared to CTDIvol, SSDEdeff has significantly improved the accuracy of dose estimation [20-22]. One of the advanced features of SSDEdeff lies in its simplicity and efficiency. AP and LAT required by deff can be easily obtained on a single axial image. However, in the anatomic region of the considerable x-ray attenuation inhomogeneity, SSDEdeff may result in misestimated radiation dose, changing with tissues attenuation characteristics [11]. A recent series of studies revealed that SSDEdeff estimated radiation dose was markedly smaller than actual patient chest absorbed dose [12,15,18]. The findings in the current study demonstrated that SSDEdeff was different from SSDEdw, with an average underestimation of 10.08% (range −2.89% to 24.19%) in CTCA. Chest is fully filled with air, which has extremely weak x-ray attenuation and much lower CT value than water. Therefore, these previous studies indicated that air was the primary factor affecting the estimation performance of SSDEdeff in chest. Contrary to these studies, Signalhigh, rather than Arealow, Areahigh, and Signallow, significantly affects the change of SSDEdeff in CTCA, and a negative relationship was found between Signalhigh and SSDEdeff in the current work. Thus, it would be theoretically expected that SSDEdeff tends to get close and even equal to SSDEdw, as Signalhigh decreases. However, intraluminal attenuation is required to meet diagnostic image quality in CTCA. The assumption that SSDEdeff is equal to SSDEdw will not be established, and difference between them will be maintained in radiologic practice.

In this study, deff was not in accordance with dw. The x-ray attenuation of air, bony and enhanced structures was considerably different from that of water. The air decreased the attenuation of patient considerably, which mainly increased the geometrical dimension. On the contrary, high x-ray attenuation bony and enhanced structures mainly resulted in increased dw. However, Arealow was significantly greater than Areahigh, while in terms of CT value, air was at the bottom level in all tissues of the scan region of CTCA. Thus, air is significantly different from bony and is enhanced in area and x-ray attenuation. It may result in 64.34% of variation in dw (R2=0.6434) explained by deff and difference between SSDEdeff and SSDEdw.

Increase in both SSDEdeff and SSDEdw with patient dw size was observed in this work. This was expected due to adjustment of scan parameters for the inter-patient acceptable diagnostic image quality. Large patient size indicates larger geometrical dimension and higher x-ray attenuation, which can cause increased visual noise, obscured anatomic details and decreased contrast to noise ratio (CNR) [23]. Thus, to maintain a comparable diagnostic image quality, larger patients are required to use more x-photon than small patients. It is noteworthy that there was no statistical significance in correlation of SSDEdw and dw. It was considered that normalized CTDIvol using dw, which combined geometrical dimension with x-ray attenuation [10,11,16], resulted in SSDEdw with less variation compared to SSDEdeff across all patients. Thus, SSDEdw was considered be a more reasonable metric to establish CT diagnostic reference level, from which patients would benefit more. On the other hand, according to the inverse exponential correlation of f and body size [12,13], small patients would be exposed to higher SSDE, large patients would be exposed to lower SSDE with the constant CTDIvol. The effect would be the same for both SSDEdeff and SSDEdw.

It was observed that there was an average Erssde of 10.48±4.76% between SSDEdeff and SSDEdw, ranging from 0.33% to 24.16%. Erssde decreased with increasing dw. It would be expected that SSDEdeff was very close to SSDEdw for larger patient. When the patient size increased beyond a certain value, SSDEdeff would equal to SSDEdw. On the contrary, when patient size shifted to the smaller end, Erssde became greater, and SSDEdeff would considerably move away from SSDEdw, which is explained by the negative exponential correlation of f and body size [12,13]. Based on the aforementioned observation no significant correlation was found between SSDEdw and dw, the analysis using SSDEdw seemed to be more beneficial for thin patients in CTCA, although SSDEdeff and SSDEdw provided the radiation dose measurements. The metric of SSDEdeff is suitable for estimating larger patient radiation dose in CTCA.

To further explore the causes of estimation Erssde, multiple stepwise regression analysis revealed that low attenuation tissues had a noticeable impact on Prosize. Combined with the positive correlation of Prosize and Erssde, it was considered that Arealow may result in the variation of Erssde, which indicated that SSDEdeff was comparable to SSDEdw with decreasing Arealow, and Signallow may result in increased Erssde to a certain extent with decreasing Signallow, which indicated the shift of SSDEdeff from SSDEdw. In clinical practice, Arealow may vary considerably from patient to patient, generally Signallow is maintained at a relatively constant level. In fact, Arealow would be the critical variable impacting on Erssde. With respect to high attenuation tissues, both Areahigh and Signalhigh did not impact on Prosize significantly, and their impact on Erssde was negligible. It was assumed that high attenuation tissues would theoretically become the key variables to impact the Prosize and Erssde with increasing Areahigh and Signalhigh. As a matter of fact, Areahigh changed within a relative narrower range from 45.06 cm2 to 100.78 cm2 contrast to the variation range of Arealow over all the patients in this work, and CT value of 300 HU was enough to ensure that the lesion could be detected efficiently, over enhanced intraluminal attenuation would cause inverse effect to obscure diagnostic performance of CCTA [24]. Thus, the probability of high attenuation tissues to significantly change Prosize and Erssde would be low in CTCA.

This study has several limitations. Firstly, the axial scan mode of fixed tube current was used to perform CTCA, which may limit the generalizability of results to the mode of automatic tube current modulation. To the best of our knowledge, the study, however, is the first report on differences between SSDEdeff and SSDEdw in CTCA. Secondly, the data used in this study was retrieved from one institution. Although standard operation procedure can be put into radiologic practice regardless of experiment and expertise variation of technologist in individual institution, it may be necessary that the suggestions of this study would be reconfirmed using multicenter dataset in future. Thirdly, dw was automatically calculated, in contrast, measurement of deff was performed manually. Thus, individual approach might result in discrepancy of body size measurements from actual values which may partially cause a bias in retrospective CT radiation dose analyses.

Conclusions

In conclusion, although both SSDEdw and SSDEdeff can be used as the radiation dose metrics in CTCA for adult patients, SSDEdeff underestimates the radiation dose by an average of about 10.08% compared to SSDEdw. The ratio of effective diameter to water equivalent diameter, especially low attenuation details in terms of area and signal intensity, had a significant effect on Erssde between SSDEdw and SSDEdeff. Therefore, SSDEdw, rather than SSDEdeff, is a relatively reasonable metric to accurately determine the radiation dose absorbed by patients in CTCA and was recommended to implement into clinical practices.