Academic Journal
Προβολή σε EDS

Compact and efficient linear accelerator phase space modeling via implicit neural representation learning.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Compact and efficient linear accelerator phase space modeling via implicit neural representation learning.
Συγγραφείς: Heng VJ; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America., Charyyev S; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America., Yang Y; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America., Chuang C; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America., Xing L; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America., Liu L; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.
Πηγή: Physics in medicine and biology [Phys Med Biol] 2026 Sep 08; Vol. 71 (17). Date of Electronic Publication: 2026 Sep 08.
Τύπος έκδοσης: Journal Article
Γλώσσα: English
Στοιχεία περιοδικού: Publisher: IOP Publishing Country of Publication: England NLM ID: 0401220 Publication Model: Electronic Cited Medium: Internet ISSN: 1361-6560 (Electronic) Linking ISSN: 00319155 NLM ISO Abbreviation: Phys Med Biol Subsets: MEDLINE
Imprint Name(s): Original Publication: Bristol : IOP Publishing
Ιατρικοί όροι (MeSH): Particle Accelerators*, Monte Carlo Method ; Representation Machine Learning
Περίληψη: Objective.Phase space files (PSFs) are commonly used to restart Monte Carlo simulations from pre-calculated conditions, but are large and unwieldy. This study introduces and evaluates a compact representation of PSF using sinusoidal implicit representation networks (SIRENs) that enables limitless sampling while preserving dosimetric accuracy.Approach.Three models were developed to account for the different types of particles found in photon phase spaces: primary photons, secondary photons, electrons. For each particle type, a sequence of 2 or 3 SIRENs were trained to learn the implicit neural representation of the probabilistic distribution of particle position, direction and energy. Once trained, PSFs of any desired size can be generated by sampling the probability density functions implicitly represented by the model. We applied this approach to vendor-provided 6 and 10 MV photon phase spaces. Model accuracy was validated by using the network-sampled particles as input to an EGSnrc Monte Carlo dose calculation in a water phantom. Scalability is further investigated by training the model on 20% of the original PSF particles.Main results.SIREN reproduced marginal probability distributions and depth dose curves with excellent agreement. Lateral profiles showed small discrepancies in penumbra shoulders. 2%/0 mm Gamma pass rates were 97.2%, 98.1%, 96.9% and 99.2% for 6X, 6FFF, 10X and 10FFF, respectively. Mean voxel-wise dose difference was 0.44% (6X), 0.36% (6FFF), 0.51% (10X) and 0.25% (10FFF) versus vendor PSFs. Training on 20% of particles slightly reduced the 6X gamma pass rate from 97.2% to 96.3% while maintaining dose profile agreement. The approach achieved30 000-fold compression (1.6 MB vs 50 GB) with an average sampling time of0.8 s perparticles on an NVIDIA RTX 6000 Ada.Significance.The SIREN-based approach allows for a compact representation of Monte Carlo PSF with high dosimetric fidelity. It remains robust with reduced training data and would enable on-the-fly sampling of particles without recycling.
(© 2026 Institute of Physics and Engineering in Medicine. All rights, including for text and data mining, AI training, and similar technologies, are reserved.)
Contributed Indexing: Keywords: Monte Carlo dose calculation; Monte Carlo phase space; implicit neural representation; sinusoidal implicit representation network
Entry Date(s): Date Created: 20260825 Date Completed: 20260908 Latest Revision: 20260908
Update Code: 20260908
DOI: 10.1088/1361-6560/ae9e7e
PMID: 42641660
Βάση Δεδομένων: MEDLINE
FullText Text:
  Availability: 0
Header DbId: cmedm
DbLabel: MEDLINE
An: 42641660
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Compact and efficient linear accelerator phase space modeling via implicit neural representation learning.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AU" term="%22Heng+VJ%22">Heng VJ</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.<br /><searchLink fieldCode="AU" term="%22Charyyev+S%22">Charyyev S</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.<br /><searchLink fieldCode="AU" term="%22Yang+Y%22">Yang Y</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.<br /><searchLink fieldCode="AU" term="%22Chuang+C%22">Chuang C</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.<br /><searchLink fieldCode="AU" term="%22Xing+L%22">Xing L</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.<br /><searchLink fieldCode="AU" term="%22Liu+L%22">Liu L</searchLink>; Department of Radiation Oncology, Stanford University, Palo Alto, CA, United States of America.
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%220401220%22">Physics in medicine and biology</searchLink> [Phys Med Biol] 2026 Sep 08; Vol. 71 (17). <i>Date of Electronic Publication: </i>2026 Sep 08.
– Name: TypePub
  Label: Publication Type
  Group: TypPub
  Data: Journal Article
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: TitleSource
  Label: Journal Info
  Group: Src
  Data: <i>Publisher: </i><searchLink fieldCode="PB" term="%22IOP+Publishing%22">IOP Publishing </searchLink><i>Country of Publication: </i>England <i>NLM ID: </i>0401220 <i>Publication Model: </i>Electronic <i>Cited Medium: </i>Internet <i>ISSN: </i>1361-6560 (Electronic) <i>Linking ISSN: </i><searchLink fieldCode="IS" term="%2200319155%22">00319155 </searchLink><i>NLM ISO Abbreviation: </i>Phys Med Biol <i>Subsets: </i>MEDLINE
– Name: PublisherInfo
  Label: Imprint Name(s)
  Group: PubInfo
  Data: <i>Original Publication</i>: Bristol : IOP Publishing
– Name: SubjectMESH
  Label: MeSH Terms
  Group: Su
  Data: <searchLink fieldCode="MM" term="%22Particle+Accelerators%22">Particle Accelerators*</searchLink><br /><searchLink fieldCode="MH" term="%22Monte+Carlo+Method%22">Monte Carlo Method</searchLink> ; <searchLink fieldCode="MH" term="%22Representation+Machine+Learning%22">Representation Machine Learning</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Objective.Phase space files (PSFs) are commonly used to restart Monte Carlo simulations from pre-calculated conditions, but are large and unwieldy. This study introduces and evaluates a compact representation of PSF using sinusoidal implicit representation networks (SIRENs) that enables limitless sampling while preserving dosimetric accuracy.Approach.Three models were developed to account for the different types of particles found in photon phase spaces: primary photons, secondary photons, electrons. For each particle type, a sequence of 2 or 3 SIRENs were trained to learn the implicit neural representation of the probabilistic distribution of particle position, direction and energy. Once trained, PSFs of any desired size can be generated by sampling the probability density functions implicitly represented by the model. We applied this approach to vendor-provided 6 and 10 MV photon phase spaces. Model accuracy was validated by using the network-sampled particles as input to an EGSnrc Monte Carlo dose calculation in a water phantom. Scalability is further investigated by training the model on 20% of the original PSF particles.Main results.SIREN reproduced marginal probability distributions and depth dose curves with excellent agreement. Lateral profiles showed small discrepancies in penumbra shoulders. 2%/0 mm Gamma pass rates were 97.2%, 98.1%, 96.9% and 99.2% for 6X, 6FFF, 10X and 10FFF, respectively. Mean voxel-wise dose difference was 0.44% (6X), 0.36% (6FFF), 0.51% (10X) and 0.25% (10FFF) versus vendor PSFs. Training on 20% of particles slightly reduced the 6X gamma pass rate from 97.2% to 96.3% while maintaining dose profile agreement. The approach achieved30 000-fold compression (1.6 MB vs 50 GB) with an average sampling time of0.8 s perparticles on an NVIDIA RTX 6000 Ada.Significance.The SIREN-based approach allows for a compact representation of Monte Carlo PSF with high dosimetric fidelity. It remains robust with reduced training data and would enable on-the-fly sampling of particles without recycling.<br /> (© 2026 Institute of Physics and Engineering in Medicine. All rights, including for text and data mining, AI training, and similar technologies, are reserved.)
– Name: SubjectMinor
  Label: Contributed Indexing
  Group:
  Data: <i>Keywords: </i>Monte Carlo dose calculation; Monte Carlo phase space; implicit neural representation; sinusoidal implicit representation network
– Name: DateEntry
  Label: Entry Date(s)
  Group: Date
  Data: <i>Date Created: </i>20260825 <i>Date Completed: </i>20260908 <i>Latest Revision: </i>20260908
– Name: DateUpdate
  Label: Update Code
  Group: Date
  Data: 20260908
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1088/1361-6560/ae9e7e
– Name: AN
  Label: PMID
  Group: ID
  Data: 42641660
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=cmedm&AN=42641660
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1088/1361-6560/ae9e7e
    Languages:
      – Code: eng
        Text: English
    Subjects:
      – SubjectFull: Monte Carlo Method
        Type: general
      – SubjectFull: Representation Machine Learning
        Type: general
      – SubjectFull: Particle Accelerators
        Type: general
    Titles:
      – TitleFull: Compact and efficient linear accelerator phase space modeling via implicit neural representation learning.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Heng VJ
      – PersonEntity:
          Name:
            NameFull: Charyyev S
      – PersonEntity:
          Name:
            NameFull: Yang Y
      – PersonEntity:
          Name:
            NameFull: Chuang C
      – PersonEntity:
          Name:
            NameFull: Xing L
      – PersonEntity:
          Name:
            NameFull: Liu L
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 08
              M: 09
              Text: 2026 Sep 08
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-electronic
              Value: 1361-6560
          Numbering:
            – Type: volume
              Value: 71
            – Type: issue
              Value: 17
          Titles:
            – TitleFull: Physics in medicine and biology
              Type: main
ResultId 1