Page 86 - Organ motion in children for high-precision radiotherapy - Sophie Huijskens

Diaphragm Tracking
In four consecutive steps, we extracted the respiratory-induced diaphragm motion from CBCTs using an adapted version of the AS method [32] (example shown in Supplementary Figure 5.1). First, a region of interest (ROI) was defined on a single projection image, including the top of the right diaphragm. For each projection image, within the selected ROI, the derivative of the grey values along the cranial- caudal (CC) direction was calculated and the resulting pixel values along each line (perpendicular to the CC direction) were summed, creating a one-dimensional image. Accumulating all one-dimensional images created a two-dimensional AS image. Along the horizontal axis of this image, representing the projection images, we manually selected the projection images corresponding to end-exhale and end- inhale positions of the diaphragm. In each of those selected projection images, we then manually determined the CC position of the diaphragm dome top. Subsequently, the pixel coordinate corresponding to the position of the diaphragm dome top was translated to millimetres relative to the patients’ planned isocenter and was also corrected for the geometry of the CBCT scanner [31]. This resulted in a patient- and CBCT-dependent timeframe describing the CC position of the diaphragm in end-exhale and end-inhale phases (peaks) over the course of CBCT acquisition.
Respiratory analysis
Details concerning the extraction of the respiratory-induced diaphragm motion characteristics and analysis thereof are presented in a schematic overview (Supplementary Figure 5.1). The amplitude was defined as the difference between the diaphragm position in the end-inhale and end-exhale phase. The cycle time equals inspiratory time plus expiratory time. Results were analysed per fraction and per patient, in order to evaluate the outcomes over the whole patient group.
For each fraction, we calculated the mean amplitude and intrafractional variability (the standard deviation (SD) over all amplitudes). For each patient, we calculated the mean amplitude, interfractional variability (i.e., the SD over mean amplitudes from each fraction), and the intrafractional variability (root mean square of the SD from each fraction). For the whole patient group, we calculated the group mean amplitude by averaging the patients’ mean amplitude, the group interfractional variability by averaging the patients’ interfractional variabilities, and the group intrafractional variability by averaging the patients’ intrafractional variabilities. Also, interpatient variability was calculated, expressed as the SD over all patients’ mean amplitude. All of this was also computed for the cycle time.
Statistical Analysis
To test for normality of our data we used the Shapiro-Wilks test. Since not all data fitted the normal distribution, we used the Spearman’s correlation test (significance level p < 0.05) to test for possible relationships between respiratory-induced diaphragm motion characteristics and patient-specific factors (age, height, and weight). We tested if respiratory-induced diaphragm motion characteristics of the patients treated under GA (n=7, age range 2-11 years) differed from patients treated without GA in a similar age range (n=12, age range 3-10 years) (Mann-Whitney-U test, significance level p<0.05).
83