Estimation and clinical verification of the effective and skin doses for pediatric and adult patients undergoing the cardiac interventional examination using five PMMA phantoms and TLD/ionization chamber technique

2.1Effective dose

ICRP committees have been quantifying personal radiation dose for several decades. According to the ICRP-60 report published in 1990, the protection quantity for personal dosimetry is termed the effective dose, E, and is defined as

where ωT denotes the weighting factor of particular tissue or organ, while HT represents the equivalent dose received by an tissue or organ, which is defined as

Here ωr is the weighting factor of the incident radiation, while DT,r denotes the mean dose of incident radiation of γ-type absorbed by the organ or tissue T [11]. The effective dose can be directly determined by the derivation of the equivalent dose for each organ or tissue and further multiplication of the obtained values by the corresponding weighting factors (cf. Eq. (2)). In accordance with ICRP-26, ωT values were explicitly assigned only to six organs, while five more organs receiving the next highest dose equivalents were jointly analyzed as a pseudo-organ called the “remainder” [12]. The provisions of ICRP-60 envisaged the allocation of ωT values to more organs than ICRP-26, as well as the refinement of available values of ωT, in view of more accumulated data concerning the risk of cancer in these organs due to the radiation exposure. The pseudo-organ or “remainder” was also re-defined in ICRP-60 and involved ten specific organs. The values ωT derived for particular organs were found to be independent of the equivalent dose HT delivery; therefore, the ICRP-26 definition of HE can be extended to include updated values of ICRP-60 ωT, yielding a more accurate estimate of E.

Figure 1.

(A) Five PMMA phantoms were put side by side with Rando phantom to demonstrate the relative geometrical size, (B) each phantom was assembled from 31 plates of various sizes and numbered sequentially from top to bottom as follows: Nos. 1–6 corresponded to head, 7–8 to neck, 9–21 to chest and abdomen, and 22–31 to pelvis.

2.2Five PMMA phantoms

Five PMMA phantoms (10, 30, 50, 70, and 90 kg, respectively) were customized to represent a baby, child, adult female, adult male, and overweight adult (obesity case), in accordance with ICRU-48 report, as shown in Fig. 1 [9]. Each heterogeneous phantom was assembled from 31 acrylic (PMMA) plates. In addition, the skull, ribs, spine, and pelvis were made of pure aluminum, while the lung was made of high-density polyethylene foamed cotton. Several through holes were drilled into each plate, then 3–5 TLD chips were inserted in these through holes for the dose survey, while acrylic plugs were used to fill empty ones. As shown in Fig. 1A, five PMMA phantoms were put side by side with Rando phantom to demonstrate the relative geometrical size, (B) each phantom was assembled from 31 plates of various sizes and numbered sequentially from top to bottom as follows: Nos. 1–6 corresponded to head, 7–8 to neck, 9–21 to chest and abdomen, and 22–31 to pelvis. The dimensions of five phantoms were designed according to the physical dimensions of linearity, i.e., weight W is directly proportional to the cubed linear dimension L(W≈L3). Therefore, the body mass index (BMI) was 13.52, 16.46, 21.64, 24.22, and 27.78 kg m-2 for the phantoms of 10, 30, 50, 70, and 90 kg, respectively. Noteworthy is that not only the arrangement of each plate but also the structure of five phantoms were identical to ensure the consistence and varied only by size. Thus, different exposed doses obtained via either X-ray or fluoroscopy can be only correlated to phantoms of different size.

2.3TLD/ionization chamber setup

One hundred five standard TLD-100 chips consisting of Lithium Fluoride (LiF: Mg, Ti; 3.2 × 3.2 × 0.89 mm3) were obtained from Thermo Fisher Scientific Inc. (formerly Harshaw, Bicron). The TLDs were randomly categorized into 35 packs, each consisting of three TLDs. The TLD assigned dose was calibrated using the Advanced Markus ionization chamber, namely Victoreen Model 4000 with a parallel plate ionization chamber. The irradiated TLD reading was obtained using a Mikro Lab RA94 TLD reader/analyzer and by annealing in furnaces (Barnstead Int. Co., model 47900) coupled with an oven/incubator (model 19200) for 400∘C 1 h and 100∘ C 2 h. Furthermore, each TLD was cooled for at least 24 hours before the next exposure, in order to efficiently suppress the residual dose. The TLD signal fading and self-absorption of TLD light can be treated as an internal interfering factor and given a negligible contribution, insofar as the exposed TLDs were always read in the same period of cooling time.

An additional high-sensitivity pencil-type ionization chamber (Victoreen, model 6000–100, 3.2 cc) was preset for measuring the fluoroscopy, since TLD had comparatively low detecting efficiency for fluoroscopy. In doing so, an auxiliary PMMA plate was specially made with 3–5 through tunnels. The pencil-type ionization chamber could be inserted into any tunnel to survey the dose, whereas acrylic sticks of the same size were inserted into the remaining 2–4 through tunnels to maintain the plate integrity, yet, the auxiliary plate could be inserted into any layer to replace the original one of the phantom. The pencil-type ionization chamber had a high sensitivity to measure the low-energy fluoroscopy and maintain the dose-response linearity, its only drawback being the inconvenience of its application and larger time required for the dose measurement. The pencil-type ionization chamber could record a single fluoroscopy dose at a time to obtain a single dose position inside the phantom. In contrast, 105 TLDs could be fully inserted into the phantom and then exposed only once to acquire the full empirical data for a single specific task.

The phantom was assembled with TLD chips or pencil-type ionization chamber inside, then exposed to the bi-plane X-ray facility that was specifically designed for the cardiac interventional examination (Philips Integris Allura 9 Biplane system) located at the Cardiac Catheterization Laboratory, Taichung Veteran General Hospital, Taiwan (CCL, TVGH). The X-ray and fluoroscopy settings for each phantom were then applied to real patients that underwent the cardiac interventional examination as listed in Table 1. Here 10- and 30-kg phantoms corresponded to baby or child, respectively, from the pediatric viewpoint. Similarly, 50-, 70-, and 90-kg phantoms were used to simulate the adult female, male, and overweight adult, respectively. The recorded DAP could be manipulated by changing the exposure time. Table 2 implies the precise arrangement of well-packed TLDs inside the phantom for X-ray exposure. Each TLD pack had three chips sealed in a PE bag, which were then inserted into specific through holes for the exposure. Also, the tissue weighting factor (cf. Eq. (1), ωT) was also listed for reference; the factor was normalized from the original ICRP-60 report to ensure its unity. Figure 2 reveals part of the data acquisition process inside the CCL. As seen in Fig. 2A, a 70 kg phantom was placed between two X-ray emitters, and the focal spots aiming at the back of phantom, (B) a 10 kg phantom, the photo was taken from the opposite side of (A), (C) five auxiliary plates with 3–5 through tunnels to insert the pencil-type ionization chamber for surveying the fluoroscopy dose. The size of an auxiliary plate equaled by the geometrical size to the respective phantom, which allowed one to insert it into any layers to replace the original one, and (D) a 90 kg phantom with auxiliary plate replacing the original 14th plate to measure the fluoroscopy dose; as is seen, the acrylic stick was extracted from the tunnel, in order to put in the pencil-type ionization chamber from another side, whereas other four through tunnels were filled with sticks to maintain the plate integrity.

Figure 2.

(A) A 70 kg phantom was placed between two X-ray emitters, and the focal spots aiming at the back of phantom, (B) a 10 kg phantom, the photo was taken from the opposite side of (A), (C) five auxiliary plate with 3-5 through holes to insert the pencil-type ionization chamber for surveying the fluoroscopy dose. The size of an auxiliary plate equaled by the geometrical size to the respective phantom, which allowed one to insert it into any layers to replace the original one, and (D) a 90 kg phantom with auxiliary plate replacing the original 14th plate to measure the fluoroscopy dose; as is seen, the acrylic plug was extracted from the hole, in order to put in the pencil-type ionization chamber from another side, whereas other four through holes were filled with plugs to maintain the plate integrity.

3.1Effective and skin doses

Table 3 lists the relevant data derived in this work. The reported data for each specific case were averaged from three independent measurements. The TLDs were measured and then averaged to represent the particular assigned organ or tissue. The data were categorized by BMI and exposed time Er [sec] (cf. Table 3), then the time was converted to different DAP of X-ray or fluoroscopy, in compliance with a preliminary survey that was calibrated using both BMI and exposed time [13]. Figure 3 depicts four correlations between DAP and effective/skin dose for five different phantom weights. Thus, a thorough survey of effective or skin doses for five phantoms exposed to X-ray or fluoroscopy was conducted. The high linearity or consistence among five correlation plots revealed a quite robust fit with a negligible systematic error.

Figure 3.

Four correlations between DAP and dose using five phantoms of different weight. Thus, a thorough survey of effective or skin doses for five phantoms exposed to X-ray or fluoroscopy was accomplished in this work. (A) Effective dose from X-ray exposure, (B) skin dose from X-ray exposure, (C) effective dose from fluoroscopy, and (D) skin dose from fluoroscopy.

3.2Error treatment

The errors associated with the derived effective or skin doses for various X-ray/fluoroscopy exposure arrangements were calculated as the square root of the sum of squared individual errors, Δ⁢i, as listed in Table 4. The uncertainty for ωT value was set to 5%, since the weighting factor was normalized in this specific measurement. The error for X-ray power fluctuation was based on the monthly clinical quality assurance (QA) at Taichung Veteran Hospital, while the internal normalization errors for TLDs were quoted from the preliminary survey of each specific TLD, as listed in Table 4 [14, 15]. The uncertainty resulting from non-tissue equivalence effect in the acrylic phantom was set to 5%, since the heterogeneous phantom was manufactured from PMMA, aluminum, and high-density polyethylene foamed cotton. Eventually, the total error as indicated in this study was mainly related to the statistical error in counting, which could be efficiently suppressed by repeated measurements. Therefore, all reported data were averaged from three independent trials, and the maximum counting statistical and total errors amounted to 6.2% and 15.4%, respectively.

4.1Optimal-fitting of a semi-empirical formula via STATISTICA program

The correlation of DAP versus dose for various phantom weights (reduced to BMI) can be expressed by a semi-empirical formula defined as a binary quadratic equation to predict the exposure dose (effective or skin one) for patients that underwent the cardiac interventional examination (via X-ray or fluoroscopy) in CCL. In doing so, the coefficients of the semi-empirical formula can be derived with the STATISTICA program [10]. The correlations among the variables are determined and defined as nonlinear models, nonlinear estimations, and user-specified regressions with customized loss functions to perform the numerical analysis using the normalized data from the real measurement. The predicted effective or skin doses are the expectation values of the computational results. Therefore, five independent groups of 25 individual datasets each [5 × 5 = 25] were incorporated into the model to optimize the compromised solution of the predicted dose. Also, five terms, including one constant, were used in the binary quadratic equation to reveal the best correlation among the variables as listed below:

Figure 4.

Plot of the expectation values predicted via the semi-empirical formulas, which was automatically constructed by the STATISTICA default feature. (A) effective dose from X-ray exposure, (B) skin dose from X-ray exposure, (C) effective dose from fluoroscopy, and (D) skin dose from fluoroscopy.

where Dose, X, and Y are the expectation values of effective or skin dose from X-ray or fluoroscopy, BMI, and DAP, respectively, whereas terms A∼E are the derived coefficients of the formula. The loss function was defined as the total deviation between the predicted and observed doses for all 25 cases, and thus, a small loss function is always preferable in the theoretical computation. Table 5 lists the derived coefficients A∼E for the four binary quadratic equations and the coefficient of determination R2, whereas the respective plot for expectation values is depicted in Fig. 4. The latter was automatically plotted by the STATISTICA default feature, and the smooth surface of the expectation values’ domain indicates the lack of conflicts among datasets and the absence of any systematic errors within the computational process. Thus, the coefficient of determination R2 reaches 0.888 ∼ 0.986, which implies a high consistence and reliable estimation for the four kinds of exposed doses in CCL. Eventually, the effective or skin doses can be easily derived by including the patient’s BMI and recorded DAP of X-ray or fluoroscopy. Also, the quadratic coefficient of DAP reaches 0.0 in all four equations (cf. Table 5). This prevents too large DAP values to confuse the estimation, since DAP values may vary by several thousand units [mGy⋅cm-2], whereas the BMI ranges only by several ten units [kg⋅m2]. However, this biased estimation can be suppressed by normalizing the respective variable to the same level and reducing its fluctuation to the range from -1.0 to +1.0 [16, 17].

4.2Clinical verification using 30 cardiac interventional patients

The obtained binary quadratic equations were applied to predict the exposed dose for 30 patients who underwent the cardiac interventional examination. This was envisaged to verify the theoretical estimation accuracy. In doing so, thirty patients were asked to place three TLDs on their backs within the X-ray focal spot aiming region during the cardiac examination. The TLD chips had to be put along the cardiac edge to avoid any false imaging being made. The TLD chips were removed immediately after the examination and followed the same process of TLD reading to collect the clinical data. Table 6 lists the derived data and their comparison with the theoretical estimation. As seen in Table 6, these results are widely fluctuated. Only 12 out of 30 cases have a disagreement lower than 100%, whereas others exhibit even worse correlation. The possible reasons of such deviations might be (A) the focal spot aiming of the X-ray is moving according to instant response from the cardiac interventional examination. In contrast, the equation is calculated according to fixed DAP that is exposed by a steady focal spot aiming, thus, the prediction might under- or overestimate the dose on the basis of inappropriate DAP index, (B) the facility can show only combined DAP index rather than individual DAP by X-ray or fluoroscopy, respectively. Thus, the DAP is roughly divided into fifty-fifty contributions of fluoroscopy or X-ray. The assumption is concluded from a long-term in-situ statistics in the CCL from May 2016 to September 2017 and suggested a “fifty-fifty” rule of thumb. The X-ray has higher specific DAP than fluoroscopy (for 70 kg phantom, X-ray and fluoroscopy exhibited approx. 630 and 72 DAP/s, respectively; whereas the time sharing between X-ray and fluoroscopy reached the ratio of 1:9 in the regular CCL diagnosis. Thus, DAP fulfills the “fifty-fifty” rule in reality. In the routine examination, the CCL staff needs to step on the foot pedal of fluoroscopy for acquiring the tentative image and then refine the preset parameters for ensuring a precise image (since the machine operates only when there is pressure on the switch); whereas the X-ray is used to ascertain the cardiac interventional process for the diagnosis or to store the printed imaging for the follow-up study. In some special cases, the fluoroscopy is continuously adopted for the benefit of low-exposure dose, and then the DAP ratio changes to “forty-sixty”. Therefore, from the statistical viewpoint, the exact breakdown of DAP between fluoroscopy and X-ray is a complicated and fuzzy process. Yet, from the clinical viewpoint, the “fifty-fifty” option is always a good compromise. For example, if a 70-kg standard male patient receives 60,000 DAP during the cardiac interventional examination, it is roughly estimated that each source shares 30,000 DAP. Then, the derived values for X-ray effective/skin dose and fluoroscopy effective/skin dose are assessed as 125, 638, 916, and 10313 μSv, respectively. If the DAP breakdown is changed to “forty-sixty” ratio, then the respective doses delivered to the patient will be 41, 116, 1150, and 12071 μSv (cf. Tabl 5). However, these predictions are based on the scenario that the focal spot experiences no shift during the examination process. Otherwise, the skin dose will be distributed along a larger area and, thus, sharply reduced.