A comparison of neural network approaches for on-line prediction in IGRT

John H. Goodband, Olivier C.L. Haas, J.A. Mills

    Research output: Contribution to journalArticle

    27 Citations (Scopus)
    24 Downloads (Pure)

    Abstract

    Image-guided radiation therapy aims to improve the accuracy of treatment delivery by tracking tumor position and compensating for observed movement. Due to system latency it is sometimes necessary to predict tumor trajectory evolution in order to facilitate changes in beam delivery. Neural networks (NNs) have previously been investigated for predicting future tumor position because of their ability to model non-linear systems. However, no attempt has been made to optimize the NN training algorithms, and no mention has been made of potential errors which can be caused by using NNs for extrapolation purposes. In this work, after giving a brief explanation of NN theory, a comparison is made between 4 different adaptive algorithms for training time-series prediction NNs. New error criteria are introduced which highlight error maxima. Results are obtained by training the NNs using previously published data. A hybrid algorithm combining Bayesian regularization with conjugate-gradient backpropagation is demonstrated to give the best average prediction accuracy, whilst a generalized regression NN is shown to reduce the possibility of isolated large prediction errors.
    Original languageEnglish
    Pages (from-to)1113
    JournalMedical Physics
    Volume35
    Issue number3
    DOIs
    Publication statusPublished - 2008

    Fingerprint

    Image-Guided Radiotherapy
    Neoplasms
    Nonlinear Dynamics
    Therapeutics

    Keywords

    • backpropagation
    • biomedical imaging
    • conjugate gradient methods
    • dosimetry
    • Gaussian processes
    • generalisation (artificial intelligence)
    • mean square error methods
    • medical computing
    • multilayer perceptrons
    • radial basis function networks
    • radiation therapy
    • regression analysis
    • tumours

    Cite this

    A comparison of neural network approaches for on-line prediction in IGRT. / Goodband, John H.; Haas, Olivier C.L.; Mills, J.A.

    In: Medical Physics, Vol. 35, No. 3, 2008, p. 1113.

    Research output: Contribution to journalArticle

    Goodband, John H. ; Haas, Olivier C.L. ; Mills, J.A. / A comparison of neural network approaches for on-line prediction in IGRT. In: Medical Physics. 2008 ; Vol. 35, No. 3. pp. 1113.
    @article{bb6f926fda9246d1ae16787f873672d6,
    title = "A comparison of neural network approaches for on-line prediction in IGRT",
    abstract = "Image-guided radiation therapy aims to improve the accuracy of treatment delivery by tracking tumor position and compensating for observed movement. Due to system latency it is sometimes necessary to predict tumor trajectory evolution in order to facilitate changes in beam delivery. Neural networks (NNs) have previously been investigated for predicting future tumor position because of their ability to model non-linear systems. However, no attempt has been made to optimize the NN training algorithms, and no mention has been made of potential errors which can be caused by using NNs for extrapolation purposes. In this work, after giving a brief explanation of NN theory, a comparison is made between 4 different adaptive algorithms for training time-series prediction NNs. New error criteria are introduced which highlight error maxima. Results are obtained by training the NNs using previously published data. A hybrid algorithm combining Bayesian regularization with conjugate-gradient backpropagation is demonstrated to give the best average prediction accuracy, whilst a generalized regression NN is shown to reduce the possibility of isolated large prediction errors.",
    keywords = "backpropagation, biomedical imaging, conjugate gradient methods, dosimetry, Gaussian processes, generalisation (artificial intelligence), mean square error methods, medical computing, multilayer perceptrons, radial basis function networks, radiation therapy, regression analysis, tumours",
    author = "Goodband, {John H.} and Haas, {Olivier C.L.} and J.A. Mills",
    year = "2008",
    doi = "10.1118/1.2836416",
    language = "English",
    volume = "35",
    pages = "1113",
    journal = "Medical Physics",
    issn = "0094-2405",
    publisher = "American Association of Physicists in Medicine",
    number = "3",

    }

    TY - JOUR

    T1 - A comparison of neural network approaches for on-line prediction in IGRT

    AU - Goodband, John H.

    AU - Haas, Olivier C.L.

    AU - Mills, J.A.

    PY - 2008

    Y1 - 2008

    N2 - Image-guided radiation therapy aims to improve the accuracy of treatment delivery by tracking tumor position and compensating for observed movement. Due to system latency it is sometimes necessary to predict tumor trajectory evolution in order to facilitate changes in beam delivery. Neural networks (NNs) have previously been investigated for predicting future tumor position because of their ability to model non-linear systems. However, no attempt has been made to optimize the NN training algorithms, and no mention has been made of potential errors which can be caused by using NNs for extrapolation purposes. In this work, after giving a brief explanation of NN theory, a comparison is made between 4 different adaptive algorithms for training time-series prediction NNs. New error criteria are introduced which highlight error maxima. Results are obtained by training the NNs using previously published data. A hybrid algorithm combining Bayesian regularization with conjugate-gradient backpropagation is demonstrated to give the best average prediction accuracy, whilst a generalized regression NN is shown to reduce the possibility of isolated large prediction errors.

    AB - Image-guided radiation therapy aims to improve the accuracy of treatment delivery by tracking tumor position and compensating for observed movement. Due to system latency it is sometimes necessary to predict tumor trajectory evolution in order to facilitate changes in beam delivery. Neural networks (NNs) have previously been investigated for predicting future tumor position because of their ability to model non-linear systems. However, no attempt has been made to optimize the NN training algorithms, and no mention has been made of potential errors which can be caused by using NNs for extrapolation purposes. In this work, after giving a brief explanation of NN theory, a comparison is made between 4 different adaptive algorithms for training time-series prediction NNs. New error criteria are introduced which highlight error maxima. Results are obtained by training the NNs using previously published data. A hybrid algorithm combining Bayesian regularization with conjugate-gradient backpropagation is demonstrated to give the best average prediction accuracy, whilst a generalized regression NN is shown to reduce the possibility of isolated large prediction errors.

    KW - backpropagation

    KW - biomedical imaging

    KW - conjugate gradient methods

    KW - dosimetry

    KW - Gaussian processes

    KW - generalisation (artificial intelligence)

    KW - mean square error methods

    KW - medical computing

    KW - multilayer perceptrons

    KW - radial basis function networks

    KW - radiation therapy

    KW - regression analysis

    KW - tumours

    U2 - 10.1118/1.2836416

    DO - 10.1118/1.2836416

    M3 - Article

    VL - 35

    SP - 1113

    JO - Medical Physics

    JF - Medical Physics

    SN - 0094-2405

    IS - 3

    ER -