Abstract

Jakob Thomsen¹, Esben Worm², Jacob Johansen¹, Per Rugaard³

¹Aarhus University Hospital, Danish Centre for Particle Therapy, Aarhus, Denmark; ²Aarhus University Hospital, Department of Medical Physics, Aarhus, Denmark; ³Aarhus University Hospital, Danish Centre for Particle Therapy , Aarhus, Denmark

Gated delivery is commonly used in radiotherapy of tumors moving with respiration. It relies on the ability to turn on the treatment beam only when the target is within a predetermined gating window. However, the latency from the target enters/exits the gating window till the beam is turned on/off affects the treatment accuracy. No standardized method for measuring the gating latency exists. We therefore propose a simple and direct method to measure the latency and demonstrate the method for a pencil-beam scanning proton system. 

A proton pencil beam was delivered at a clinical facility (ProBeam, Varian) to a 5mm cubic scintillating ZnSe:O crystal. The crystal emitted visible light when irradiated by the beam. The beam delivery was gated by optical monitoring of a marker block (RPM, Varian) on a motion stage (Fig 1.A-B). The motion stage performed vertical sinusoidal motion with 1cm peak-to-peak amplitude and 1-8s periods. The gating window was set approximately from the middle of the motion and above giving a duty cycle close to 50%. A video camera (GoPro) acquired images at 120Hz showing both the crystal and the marker block. Post-treatment the marker block position was segmented in all video frames. A sinusoidal fit provided the times T_max where the marker block was at maximum excursion. Analysis of the crystal light intensity in the video frames provided the beam-on and beam-off times for each cycle with an estimated accuracy <2ms (Fig 1C). The latencies for gate-off (t_gate-off) and gate-on (t_gate-on) were then determined as follows.

If t_gate-off and t_gate-on were zero, the mid-time of the light signal (T_light) would coincide with the time of maximum marker block excursion T_max. With finite latencies it can be shown that the light signal has a delay T_light – T_max = (t_gate-off + t_gate-on)/2. The time difference between the motion and the light signal therefore directly provided the sum of the gate-off and gate-on latency for each monitored cycle. It can furthermore be shown that the light signal duration (DT_light) increases linearly with the sinusoidal period T as follows: DT_light = aT + (t_gate-off - t_gate-on). Here, a is the constant fraction of time, where the RPM marker block is inside the gating window. A linear fit of the observed light signal duration as function of the sine period therefore provided the difference between the gate-off and gate-on latency.

The mean of the summed gating latency t_gate-off + t_gate-on across 86 analyzed cycles was 396ms (Fig 2A). The gate-on latency was longer than the gate-off latency with t_gate-off - t_gate-on = -180ms (Fig 2B). The mean gating latencies were t_gate-on = 288ms and t_gate-off = 108ms (Fig 2C).

We propose a simple and direct method to determine the gating latency through video recording of a motion stage and a scintillating crystal placed in the beam. For the investigated proton beam, the average latency was 288 ms for beam-on and 108ms for beam-off.