Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Aug 24:4:649917.
doi: 10.3389/frai.2021.649917. eCollection 2021.

Cosmic Ray Background Removal With Deep Neural Networks in SBND

R Acciarri  1 , C Adams  2 , C Andreopoulos  3   4 , J Asaadi  5 , M Babicz  6 , C Backhouse  7 , W Badgett  1 , L Bagby  1 , D Barker  8 , V Basque  9 , M C Q Bazetto  10   11 , M Betancourt  1 , A Bhanderi  9 , A Bhat  12 , C Bonifazi  13 , D Brailsford  14 , A G Brandt  5 , T Brooks  8 , M F Carneiro  15 , Y Chen  16 , H Chen  15 , G Chisnall  17 , J I Crespo-Anadón  18 , E Cristaldo  19 , C Cuesta  18 , I L de Icaza Astiz  17 , A De Roeck  6 , G de Sá Pereira  3   4 , M Del Tutto  1 , V Di Benedetto  1 , A Ereditato  16 , J J Evans  9 , A C Ezeribe  8 , R S Fitzpatrick  20 , B T Fleming  21 , W Foreman  22 , D Franco  21 , I Furic  23 , A P Furmanski  24 , S Gao  15 , D Garcia-Gamez  25 , H Frandini  10 , G Ge  26 , I Gil-Botella  18 , S Gollapinni  27   28 , O Goodwin  9 , P Green  9 , W C Griffith  17 , R Guenette  29 , P Guzowski  9 , T Ham  3 , J Henzerling  3 , A Holin  7 , B Howard  1 , R S Jones  3 , D Kalra  26 , G Karagiorgi  26 , L Kashur  30 , W Ketchum  1 , M J Kim  1 , V A Kudryavtsev  8 , J Larkin  15 , H Lay  14 , I Lepetic  31 , B R Littlejohn  22 , W C Louis  27 , A A Machado  10 , M Malek  8 , D Mardsen  9 , C Mariani  32 , F Marinho  33 , A Mastbaum  31 , K Mavrokoridis  3 , N McConkey  9 , V Meddage  23 , D P Méndez  15 , T Mettler  16 , K Mistry  9 , A Mogan  28 , J Molina  19 , M Mooney  30 , L Mora  9 , C A Moura  34 , J Mousseau  20 , A Navrer-Agasson  9 , F J Nicolas-Arnaldos  25 , J A Nowak  14 , O Palamara  1 , V Pandey  23 , J Pater  9 , L Paulucci  34 , V L Pimentel  10   11 , F Psihas  1 , G Putnam  35 , X Qian  15 , E Raguzin  15 , H Ray  23 , M Reggiani-Guzzo  9 , D Rivera  36 , M Roda  3 , M Ross-Lonergan  26 , G Scanavini  21 , A Scarff  8 , D W Schmitz  35 , A Schukraft  1 , E Segreto  10 , M Soares Nunes  12 , M Soderberg  12 , S Söldner-Rembold  9 , J Spitz  20 , N J C Spooner  8 , M Stancari  1 , G V Stenico  10 , A Szelc  9 , W Tang  28 , J Tena Vidal  3 , D Torretta  1 , M Toups  1 , C Touramanis  3 , M Tripathi  23 , S Tufanli  6 , E Tyley  8 , G A Valdiviesso  37 , E Worcester  15 , M Worcester  15 , G Yarbrough  28 , J Yu  5 , B Zamorano  25 , J Zennamo  1 , A Zglam  8
Affiliations

Cosmic Ray Background Removal With Deep Neural Networks in SBND

R Acciarri et al. Front Artif Intell. .

Abstract

In liquid argon time projection chambers exposed to neutrino beams and running on or near surface levels, cosmic muons, and other cosmic particles are incident on the detectors while a single neutrino-induced event is being recorded. In practice, this means that data from surface liquid argon time projection chambers will be dominated by cosmic particles, both as a source of event triggers and as the majority of the particle count in true neutrino-triggered events. In this work, we demonstrate a novel application of deep learning techniques to remove these background particles by applying deep learning on full detector images from the SBND detector, the near detector in the Fermilab Short-Baseline Neutrino Program. We use this technique to identify, on a pixel-by-pixel level, whether recorded activity originated from cosmic particles or neutrino interactions.

Keywords: SBN program; SBND; UNet; deep learning; liquid Ar detectors; neutrino physics.

PubMed Disclaimer

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Figures

Figure 1
Figure 1
An illustration of the SBND TPC used in this work. In this image, a neutrino interacts in the left TPC, and the outgoing particles cross the central cathode into the right TPC. The top-down projection images (vertical wire planes) are shown, which are combined into one image as seen in Figure 4.
Figure 2
Figure 2
Engineering diagram of the SBND LArTPC and its surrounding subsystems. Here, the TPC is shown lifted above the cryostat for clarity.
Figure 3
Figure 3
Neutrino energy of interactions produced for this analysis. Most neutral current events are produced by muon-type neutrinos, and so the νμ CC and Neutral Current energy spectra are similar. The relative populations here are for the dataset used in this paper, while in the neutrino beam the muon neutrino interactions are far more frequent than electron neutrino.
Figure 4
Figure 4
The raw data for one image in the dataset at full resolution. Charge observed is colored with blue for smaller charge depositions and red for larger charge depositions. There is an electron neutrino charged current interaction in the Upper TPC.
Figure 5
Figure 5
The labels for the images in Figure 4 in the dataset at full resolution. White pixels are background, gray pixels are associated with cosmic particles, and red pixels are associated with a neutrino interaction. Plane 2 shows a case of overlap between cosmic and neutrino pixels.
Figure 6
Figure 6
A representation of the multi-plane UResNet architecture. Only two of the three planes are shown in this image for clarity.
Figure 7
Figure 7
Distribution of pixel occupancies, by label, in this dataset. In general, the cosmic-labeled pixels are <1% of pixels and the neutrino-labeled pixels are <0.3%.
Figure 8
Figure 8
The training progression of the Multiplane UResNet model, trained for 25 k iterations. The light blue curve is the training performance at each step, overlaid with a smoothed representation of the same data, and a smoothed representation of the test set.
Figure 9
Figure 9
Metric performance across neutrino interaction types, as a function of neutrino energy. The solid lines are the Intersection over Union for the neutrino predicted/labeled pixels, while the dashed lines are the Intersection over Union for the cosmic predicted/labeled pixels. Each color in this plot represents the IoU for all events containing that particular neutrino interaction.
See this image and copyright information in PMC

References

    1. Abadi M., Agarwal A., Barham P., Brevdo E., Chen Z., Citro C., et al. . (2015). TensorFlow: Large-Scale Machine Learning on Heterogeneous Systems. Available online at: tensorflow.org.
    1. Abi B., Acciarri R., Acero M. A., Adamowski M., Adams C., Adams D., et al. . (2018a). The DUNE far detector interim design report volume 1: physics, technology and strategies. arXiv:1807.10334. 10.2172/1529363 - DOI
    1. Abi B., Acciarri R., Acero M. A., Adamowski M., Adams C., Adams D., et al. . (2018b). The DUNE far detector interim design report, volume 2: single-phase module. arXiv:1807.10327. 10.2172/1529362 - DOI
    1. Abi B., Acciarri R., Acero M. A., Adamowski M., Adams C., Adams D., et al. . (2018c). The DUNE far detector interim design report, volume 3: dual-phase module. arXiv:1807.10340. 10.2172/1529361 - DOI
    1. Abi B., Acciarri R., Acero M. A., Adamowski M., Adams C., Adams D. L., et al. . (2017). The Single-Phase ProtoDUNE Technical Design Report Proto DUNE. 10.2172/1366526 - DOI