In total, we scanned 9 different settings (OA: 33%, 50 % and 100%, image quality: +, 0, -) for each spine specimen (N = 4) resulting in 36 examinations with digital fluoroscopy. The automatic image generation worked in all 36 examinations. On average, 24.6 images (standard deviation +/- 5.6) were generated per examination. The mean value of the DAP over all examinations was 0.70 µGy·m² with a standard deviation of ± 0.30 µGy·m².
For the comparison images with digital X-ray we found a mean value of DAP of 30.32 µGy·m² with a standard deviation of ± 1.1 µGy·m².
First we asked if there is a difference in DAP depending on the opening area and image quality setting. We therefore analyzed the 36 measurements of the DAP and grouped them after the opening area setting (100%, 50% and 33%). There were no outliers in the data and the Shapiro-Wilk test showed a normal distribution of the measured values (p> 0.05 in all 3 groups) (see figure 2A). The values for the DAP increased from the small opening area (33%, Mean (M) = 0.56 µGy·m², Standard-Deviation (SD) = .17) over 50% opening area (M = 0.70 µGy·m², SD = .22) to the maximum opening area (100%; M = 0.82 µGy·m², SD = .25) by 45% (see Table 2a).
a
|
|
|
|
|
|
|
|
|
|
Standard
|
|
|
|
|
N
|
Mean DAP
|
deviation
|
Minimum
|
Maximum
|
Median
|
100%
|
12
|
0.82
|
0.26
|
0.46
|
1.36
|
0.89
|
50%
|
12
|
0.70
|
0.22
|
0.39
|
1.02
|
0.72
|
33%
|
12
|
0.56
|
0.17
|
0.30
|
0.81
|
0.59
|
|
|
|
|
|
|
|
b
|
|
|
|
|
|
|
|
|
|
Standard
|
|
|
|
|
N
|
Mean DAP
|
deviation
|
Minimum
|
Maximum
|
Median
|
-
|
12
|
0.57
|
0.22
|
0.30
|
1.00
|
0.48
|
o
|
12
|
0.67
|
0.15
|
0.47
|
1.00
|
0.63
|
+
|
12
|
0.84
|
0.26
|
0.39
|
1.36
|
0.86
|
Table 2: DAP values in µGy·m² for the settings (a) opening area and (b) image quality.
In addition, the individual pairs were compared using a paired t-test and the p-values were adjusted in accordance with the Bonferroni correction for multiple testing. Finally we calculated the effect strengths (Cohens dz) with the formula dz = (M1-M2)/SD. |d| = 0,2 means small effect, |d| = 0,5 means medium effect, |d| = 0,8 means strong effect. The first pair 50% vs. 33% OA showed significant differences with strong effect (t(11) = 9.75, p < .001, dz = 2.81), same in second pair 100 % vs. 33% (t(11) = 4.82, p = 0.003, dz = 1.30). No significant change after Bonferroni correction with medium effect size was found in 100% vs. 50% OA (t(11) = 2.28, p = 0.132, dz = 1.30) (see figure 2a).
See figure_2.jpg
Figure 2: Dose area product in µGy·m². Boxplots displaying the Dose area product [µGy·m²] measured in the different settings of fluoroscopy. (a) Opening area in%, (b) Image quality setting. The data show a clear trend: A large opening area and a + setting of the image quality. * denotes significant differences (p< 0.05) after paired t-test and Bonferroni correction for multiple testing.
Subsequently, we performed the same procedure for image quality setting and found an increase of DAP from the low image quality setting (M = 0.57 µGy·m², SD = 0.22) over the medium setting (M = 0.67 µGy·m², SD = 0.15) to the high setting (M = 0.84 µGy·m², SD = 0.26) by 48%.
The first pair o vs. + showed significant differences with strong effect (t(11) = -2.86, p = 0.045, dz = 0.85). The other pairs showed no significant differences: - vs. + (t(11) = -2.53, p = 0.084, dz = 0.73) and o vs. - (t(11) = -1.53, p = 0.465, dz = 0.45) (see figure 2 b).
In conclusion, we can state that there is an increase in DAP from small to large aperture area and from low to high image quality, even if not all differences have become significant.
Finally, we wanted to establish the optimal settings for further human studies. We therefore analyzed the 108 point scores resulting from the scoring of 36 examinations through 3 observers and grouped them by opening area setting (100%, 50% and 33%). The point score improved from 100 % OA (Median (MED) = 9.5) over 50 % OA (MED = 8) to 33% OA (MED = 7). The data showed no normal distribution in the Shapiro-Wilk test). Due to the lack of a normal distribution, we used the non-parametric Friedman´s test. The three settings (100%, 50% and 33%) showed significant changes in point score (Friedman test χ²(2) = 40.32, p < 0.001, n = 36). Subsequently the results of Friedman’s test underwent Dunn´s pairwise post hoc tests and Bonferroni adjustments of the p-values. Significant differences were found between all settings (100vs.50 p=0.01, 50vs.33 p=0.001, 100vs.33 p<0.001).
The effect sizes (r) were calculated using the formula r = z/√n (z = test statistic, n = number of pairs). A strong effect (r > 0.50) was found in 100% vs. 33% OA (r = .8) and a medium effect (0.3 ≤ r < 0.5) in 100% vs. 50% OA (r = 0.40) and 50% vs. 33% (r = .4). Next we grouped the data by image quality setting (-, o, +) and repeated the procedure. The point score improved from low image quality setting (MED = 9) over medium setting (MED = 8) to high setting (MED = 7). The Friedman test showed again significant changes in point score (χ²(2) = 22.51, p < 0.001, n = 36). Significant differences in the post hoc analyze were found between + and – setting (p < 0.001; r = 0.43), as well as + and o (p = 0.004; r = 0.55). Looking at the results of OA and image quality in combination, the setting 33/+ shows the best point scores in the image assessment (see figure 5).
See figure_5.jpg
Figure 5: Scatter plot with jittering of the different settings depending on the determined point values (Kloth score). The maximum value was 4 points (optimal assessment of the image), the minimum value 16 points (evaluation not possible).