Intrafraction displacement of prone versus supine prostate positioning monitored by real-time electromagnetic tracking

Abstract

Implanted radiofrequency transponders were used for real-time monitoring of the intrafraction prostate displacement between patients in the prone position and the same patients in the supine position. Thirteen patients had three transponders implanted transperineally and were treated prone with a custom-fitted thermoplastic immobilization device. After collecting data from the last fraction, patients were realigned in the supine position and the displacements of the transponders were monitored for 5-7 minutes. Fourier transforms were applied to the data from each patient to determine periodicity and its amplitude. To remove auto correlation from the stream of displacement data, the distribution of short-term and long-term velocity components were calculated from Poincaré plots of paired sequential vector displacements. The mean absolute displacement was significantly greater prone than supine in the superior-inferior (SI) plane (1.2 ± 0.6 mm vs. 0.6 ± 0.4 mm, p= 0.015), but not for the lateral or anterior-posterior (AP) planes. Displacements were least in the lateral direction. Fourier analyses showed the amplitude of respiratory oscillations was much greater for the SI and AP planes in the prone versus the supine position. Analysis of Poincaré plots confirmed greater short-term variance in the prone position, but no difference in the long-term variance. The centroid of the implanted transponders was offset from the treatment isocenter by > 5 mm for 1.9% of the time versus 0.8% of the time for supine. These results confirmed significantly greater net intrafraction prostate displacement of patients in the prone position than in the supine position, but most of the difference was due to respiration-induced motion that was most pronounced in the SI and AP directions. Because the respiratory motion remained within the action threshold and also within our 5 mm treatment planning margins, there is no compelling reason to choose one treatment position over the other.

Figure 1

Prostate displacement versus time as determined by the centroid of the radiofrequency transducers: (a) patient prone, lateral displacements; (b) patient supine, lateral displacements; (c) patient prone, superior–inferior displacements; (d) patient supine, superior–inferior displacements; (e) patient prone, posterior–anterior displacements; (f) patient supine, posterior–anterior displacements. The middle curve on each graph is the mean displacement of the 13 patients in the study, and the upper and lower curves are the maximum and minimum, respectively.

Figure 2

Net prostate displacement vs. time as calculated from the displacement of the centroid of the radiofrequency transducers in each Cartesian plane: (a) patient prone; (b) patient supine. The middle curve on each graph is the mean displacement of the 13 patients in the study, and the upper and lower curves are the maximum and minimum, respectively.

Figure 3

Composite Fourier transforms of displacement/time graphs into the relative amplitude/frequency domain for the study population: (a) left–right lateral; (b) superior–inferior; (c) anterior–posterior. The amplitude of the 10 Hz sampling frequency was defined as 100. In each graph, the amplitudes of respiratory oscillations in the range 0.2–0.4 Hz are much greater for patients in the prone position than when supine.

Figure 4

Poincaré plot and variance histograms of a patient in the prone position. Vector displacements were calculated for 3330 measurements of the centroid of three radiofrequency transponders in three cardinal planes. Each vector displacement in the acquisition sequence, n, was plotted against the next acquisition, $\mathrm{n}+1$. A histogram of the short‐term variance or velocity relative to the diagonal identity line passing through the origin of the scatter plot is at the lower left. At the lower right is a histogram of the long‐term velocity, which measures the offset of each point relative to the line normal to the identity line and passing through the mean vector on the scatter plot. The scale of the latter histogram is about 10 times the scale of the former.

Figure 5

Histograms of the short‐term and long‐term variance for the 13 patients in the prone and supine positions: (a) short‐term velocity, prone; (b) short‐term velocity, supine; (c) long‐term velocity, prone; (d) long‐term velocity, supine. By the nonparametric Wilcoxon matched‐pairs signed‐rank test, the short‐term distributions (a) and (b) were significantly different for the two positions; however, the long‐term distributions (c) and (d) did not differ significantly.

Figure 6

Net vector displacements (a) of the prostate for the 13 patients are correlated to systematic variations. Stratified by prone or supine positioning, only two patients (# 2 and # 5) had greater displacement in the supine position than prone. Mean vector standard deviation (b) of the vector displacements of the ~ 3,300 data points per patient. The SD of the vector displacements is correlated to random variance. Only three patients (#s 2, 5, and 12) had greater variance in the supine position than prone.