1. Abstract¶

Hadron therapy is an upcoming cancer treatment modality with a potentially better spatial accuracy than X-ray irradiation. While an X-ray beam displays a exponential decay in intensity as it traverses matter, a hadron beam has depth profile with a sharp peak, at which point most particles will be stopped. The practical consequence of this behavior is that hadrons allow dose to be deposited more accurately than X-rays. Hadron therapy is therefore well suited to the treatment of tumors located close to sensitive organs. Imaging of the tumor and its surrounding tissue still heavily relies on X-ray CT. To construct a treatment plan for hadron therapy, the X-ray radiodensity map must be converted to a stopping power map for particle beams. This conversion introduces an inherent uncertainty of up to 3%, compromising the improved accuracy that hadron therapy potentially provides. Imaging in the same modality as the treatment would remove this conversion error. This thesis describes the Nikhef/KVI Proton Imager, a device built to measure the hadron radiodensity directly.

To construct a hadron CT image, the Detector Research and Development group of Nikhef in Amsterdam designed a prototype setup with a GridPix as a particle tracker and a crystal with PMT as calorimeter. GridPix is a miniature drift chamber atop a TimePix 256x256 pixel chip, allowing particles to be tracked in 3D. In November 2011, the first test of the setup was conducted at the KVI proton accelerator, which proved that the concept works. The calorimeter and tracker data were successfully correlated to objects placed in the beam. The GridPix readout is limited to an event acquisition rate of 120Hz, but a rate of 2Hz was attained, chiefly due to a conservative trigger design and a suboptimal calorimeter data acquisition system. In order to increase the event acquisition rate to 120Hz, these two parts of the setup were overhauled. A new calorimeter data acquisition system can read out continuously and indefinitely, greatly enhancing the previous run length of 256 events. Readout rates up to 31kHz were obtained. A new trigger design takes into account a busy signal from the GridPix readout system, and reduces a timeout of 500ms to the approximately 10ms GridPix readout time, increasing the potential event acquisition rate to at least 100Hz. The final systems were tested in a multiday run measuring cosmics and in a short run with a frequency generator. A second test at the KVI proton accelerator should confirm the performance improvements, but has not yet been conducted.