Development of a proton Computed Tomography (pCT) scanner at NIU
S. A. Uzunyan, G. Blazey, S. Boi, G. Coutrakon, A. Dyshkant, B. Erdelyi, A. Gearhart, D. Hedin, E. Johnson, J. Krider, V. Zutshi, R. Ford, T. Fitzpatrick, G. Sellberg, J. E. Rauch, M. Roman, P. Rubinov, P. Wilson, K. Lalwani, M. Naimuddin
aa r X i v : . [ phy s i c s . i n s - d e t ] D ec Proceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013
Development of a proton ComputedTomography (pCT) scanner at NIU
S. A. Uzunyan , G. Blazey , S. Boi , G. Coutrakon ,A. Dyshkant , B. Erdelyi , A. Gearhart , D. Hedin ,E. Johnson , J. Krider , V. Zutshi , R. Ford ,T. Fitzpatrick , G. Sellberg , J. E. Rauch , M. Roman ,P. Rubinov , P. Wilson , K. Lalwani , M. Naimuddin Dept. of Physics, Northern Illinois University, DeKalb, IL 60115, USA Fermi National Accelerator Laboratory, Batavia, IL 60510, USA Delhi University, 110007, India
Abstract
We describe the development of a proton Computed Tomography(pCT) scanner at Northern Illinois University (NIU) in collabo-ration with Fermilab and Delhi University. This paper providesan overview of major components of the scanner and a detaileddescription of the data acquisition system (DAQ). roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 Images with protons provide electron density along the proton path in thebody of a patient. The electron density determines the penetration range fora proton of a certain energy, thereby allowing accurate location of the Braggpeak inside a tumor volume. Proton imaging can provide range uncertaintiesof about 1% compared to 3-4% achievable via traditional X-ray computedtomography, while also inducing a lower dose for image production [1]. Todate a prototype scanner capable of producing images of the required qualitywas built at Loma Linda University Medical Center (LLUMC) in 2010 [2].The pCT Phase II scanner constructed at Northern Illinois University (NIU)is a successor of this device. It is designed to demonstrate pCT can beused in a clinical environment and has the ability to collect data requiredfor 2D or 3D image reconstruction in less than 10 min. We concentratehere on the data acquisition system. The detailed description of the scannerhardware components is given in [3], and the image reconstruction hardwareand software are described in [4].
The scanner side view is shown in Figure 1, corresponding to the geome-try used for the detector simulation in GEANT [5]. The key elements arethe fiber tracker (FT) consisting of four X-Y stations (spatial resolution of ∼ √
12) before and after a rotating Head Phantom, and the range de-tector, a calorimeter stack consisting of 96, 3.2 mm thick, scintillating tiles.The signal readout in both detectors ( ∼ ≈ × proton historiesfor one 3D image of a human head at a data collection rate of 2 MHz orfaster. 2 roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 Figure 1: A schematic of the NIU Phase II pCT detector.
The SiPMs signals from the fiber tracker planes and from the calorimeterstack are collected and digitized by the 16 or 32 channel FPGA-based front-end electronics boards. The boards send digitized data to the DAQ systemthrough 20 UDP streams (eight are reserved for the fiber tracker and 12 forthe calorimeter stack) over 1 Gbit/s ethernet connections. There is no anexternal trigger: each board reads out all of its channels if at least one ofthem has a signal above a threshhold. The data are shipped in the followingformats: • fiber tracker RAW data. The fibers in the fiber tracker planes are bundledin groups of three neighbor fibers. This design allows the incident protonto simultaneously hit two adjacent bundles and thus the front-end reportspaired hits: the local bundle number ( lbn ) of the first bundle in a pair andthe state (fired or not) of the ( lbn +1) neighbor. The timestamp ( ts ) is addedto distinguish hits of different proton histories. • calorimeter stack RAW data. The scintillator planes in the calorimeterstack are grouped in eight. For each group the front-end reports: the planenumber LP max with the maximum energy deposition, the amplitude A max ofthis maximum energy deposition, the fractional (to the A max ) amplitudes inthe remaining seven planes, and the timestamp.The size of the fiber tracker and calorimeter hits in the described design isthree and six bytes, respectively. At the readout rate of 2 MHz this requires3 roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 × histories we expect a 208 GBRAW data sample. (a) (b) Figure 2: The bit content of RAW (input) events from the (a) fiber trackerand (b) calorimeter front-end channels.
The complete DAQ system, shown in Figure 3, was assembled and commis-sioned in January-March 2013. The six worker nodes and the head node forma cluster that provides 24 input channels to collect front-end data, 72 CPUcores (running at 2.6 GHz) for the data processing, and 9 TB disk storagespace. The head node runs cluster management software and is remotelyaccessible from an operator desktop. All nodes are interconnected with a2 Gbit/s internal network. The DAQ software uses the free Scientific Linux6.2 operating system, with the event collector and processing modules de-veloped based on the ROOT [7] data analysis tools. As tested, this systemis capable of accepting data at a rate up to 50 MB/s per input stream withan error rate less than 0 . roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 records each proton track (the eight hits in the fiber planes), the rotationangle of the detector, and the energy deposited in the calorimeter stack. For2 × proton histories the 48 GB data file will be stored for subsequentimage reconstruction at the NIU Compute Cluster.Figure 3: A diagram of the DAQ software modules, the assembled DAQcluster, and the bit content of the output event. In the Fall of 2012, the DAQ reconstruction software was used for the datataking control and for the data analysis in tests of the fiber tracker andcalorimeter prototypes at LLUMC. After assembling the calorimeter thissoftware was again used for analysis of tests conducted at the ProCure Protoncenter in Warrenville, Illinois. Figure 4 shows the first results of the Braggpeak measurement for a 200 MeV proton beam.5 roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 (a) (b)
Figure 4: a) The assembled calorimeter stack at ProCure Proton center inWarrenville, Illinois; b) the average maximum of the calorimeter stack tilesignals (in ADC counts) versus tile number collected from 9000 200 MeVprotons.
The major components of the NIU Phase II pCT scanner (the calorimeter,the fiber tracker and the DAQ system) were assembled by November 2013and are being commissioned. The complete system will be tested in 2014.The detailed project documentation can be found at [8].
We thank the staffs at Fermilab and collaborating institutions, and acknowl-edge support from the US Department of Defense.
References [1] Z. Liang et al. “Proton Computed Tomography”, in M. A. Hayat, Ed.,Cancer Imaging-Instrumentation and Applications, vol.2, 99-120, 2007.[2] H.F.-W. Sadrozinski et al. , IEEE NSS-MIC Conference, 4457-4461,(2011).[3] G. Coutrakon et al. , Proceedings AccApp 2013, Bruges, Belgium.6 roceedings of the “New Trends in High Energy Physics” Conference,Alushta, Crimea, September 23-29, 2013 [4] N. T. Karonis et al.
J. Parallel Distrib. Comput., (2013).http://dx.doi.org/10.1016/j.jpdc.2013.07.016[5] A. Agostinelli et al.et al.