[Paper Review] A New Proton CT Scanner

S. A. Uzunyan, G. Blazey+16 more|arXiv (Cornell University)|Aug 29, 2014

Radiation Therapy and Dosimetry4 references5 citations

TL;DR

This paper presents a second-generation proton CT scanner designed to improve proton therapy treatment planning by directly measuring relative stopping power (RSP) with sub-1% uncertainty, reducing range uncertainties from 3–4% (using X-ray CT) to under 1%. The system uses a fiber tracker and scintillator stack detector to reconstruct proton tracks and energy loss at up to 2 MHz, enabling high-precision 3D RSP mapping of head phantoms with minimal dose and no metal artifact issues.

ABSTRACT

The design, construction, and preliminary testing of a second generation proton CT scanner is presented. All current treatment planning systems at proton therapy centers use X-ray CT as the primary imaging modality for treatment planning to calculate doses to tumor and healthy tissues. One of the limitations of X-ray CT is in the conversion of X-ray attenuation coefficients to relative (proton) stopping powers, or RSP. This results in more proton range uncertainty, larger target volumes and therefore, more dose to healthy tissues. To help improve this, we present a novel scanner capable of high dose rates, up to 2~MHz, and large area coverage, 20~x~24~cm$^2$, for imaging an adult head phantom and reconstructing more accurate RSP values.

Motivation & Objective

• Address the systematic error in X-ray CT-based proton stopping power estimation, which causes 3–4% range uncertainty in proton therapy.
• Develop a proton CT scanner capable of high data rates (up to 2 MHz) to enable clinically feasible scan times.
• Achieve high spatial resolution (0.27 mm) and low mass (WET <1 mm per tracker plane) to minimize multiple scattering and improve image accuracy.
• Enable direct measurement of water equivalent path length (WEPL) and proton trajectory for accurate 3D relative stopping power (RSP) reconstruction.
• Minimize artifacts from dental or metallic implants by replacing X-ray imaging with proton-based CT.

Proposed method

• Utilizes eight tracker planes (four upstream, four downstream) with 0.5 mm diameter scintillating fibers to measure X and Y coordinates of individual proton tracks.
• Employs a 96-tile scintillator stack (3.2 mm thick per tile) to measure residual energy and determine water equivalent path length (WEPL) for each proton.
• Applies a 'most likely path' formalism to associate proton tracks with 1 mm³ voxels in the scanned volume for image reconstruction.
• Uses front-end electronics (PAD-E) with SiPM readout, 16-bit timestamps at 75 MHz, and 1 ms data framing to enable high-speed, synchronized data acquisition.
• Employs a 1 Gbit/s Ethernet-based DAQ system with UDP protocol to stream compressed data (25–75 bytes per event) to a CPU/GPU cluster for post-processing.
• Reconstructs single proton histories by coalescing tracker and calorimeter hits within a 100 ns time window and filtering for pile-up.

Experimental results

Research questions

• RQ1 Can a proton CT scanner achieve sub-1% relative stopping power uncertainty, significantly improving upon the 3–4% uncertainty from X-ray CT?
• RQ2 Can a fiber tracker and scintillator stack detector system achieve a data rate of 2 MHz while maintaining 0.27 mm spatial resolution and low mass (WET <1 mm)?
• RQ3 Can the system reconstruct high-fidelity 3D RSP images of a head phantom with minimal artifacts and without beamline table motion?
• RQ4 Can the DAQ system handle high-rate proton events (2 million/sec) with low dead time and reliable synchronization across 2100 readout channels?
• RQ5 Can the image reconstruction pipeline accurately extract single proton histories and generate 3D RSP maps suitable for clinical proton therapy planning?

Key findings

• The NIU Phase II proton CT scanner was fully assembled and installed at Central DuPage Hospital for testing in a 200 MeV proton beam.
• The fiber tracker achieved a spatial resolution of 0.27 mm, calculated as pitch / √12, with 2100 readout channels across 8 planes.
• Signal-to-noise measurements showed 10–25 photo-electrons per proton per channel in beam spot areas, confirming sufficient signal fidelity.
• The system demonstrated data acquisition at up to 2 MHz, with data compression and 1 ms framing enabling efficient transfer via 1 Gbit/s Ethernet.
• The GPU cluster successfully reconstructed high-quality 3D images of a 14 cm diameter Lucy® QA phantom using data from a prior prototype, validating the reconstruction pipeline.
• The scanner is compatible with standard proton therapy gantry geometry and enables full 360° rotation of the head phantom in a single scan without table motion.