Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer

Anders B. Olin; Christopher Thomas; Adam E. Hansen; Jacob H. Rasmussen; Georgios Krokos; Teresa Guerrero Urbano; Andriana Michaelidou; Björn Jakoby; Claes N. Ladefoged; Anne K. Berthelsen; Katrin Håkansson; Ivan R. Vogelius; Lena Specht; Sally F. Barrington; Flemming L. Andersen; Barbara M. Fischer

doi:10.1016/j.adro.2021.100762

Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer

Anders B. Olin^*, Christopher Thomas, Adam E. Hansen, Jacob H. Rasmussen, Georgios Krokos, Teresa Guerrero Urbano, Andriana Michaelidou, Björn Jakoby, Claes N. Ladefoged, Anne K. Berthelsen, Katrin Håkansson, Ivan R. Vogelius, Lena Specht, Sally F. Barrington, Flemming L. Andersen, Barbara M. Fischer

^*Corresponding author for this work

Research output: Contribution to journal › Article › peer-review

6 Citations (Scopus)

Abstract

Purpose: Radiotherapy planning based only on positron emission tomography/magnetic resonance imaging (PET/MRI) lacks computed tomography (CT) information required for dose calculations. In this study, a previously developed deep learning model for creating synthetic CT (sCT) from MRI in patients with head and neck cancer was evaluated in 2 scenarios: (1) using an independent external dataset, and (2) using a local dataset after an update of the model related to scanner software-induced changes to the input MRI.

Methods and Materials: Six patients from an external site and 17 patients from a local cohort were analyzed separately. Each patient underwent a CT and a PET/MRI with a Dixon MRI sequence over either one (external) or 2 (local) bed positions. For the external cohort, a previously developed deep learning model for deriving sCT from Dixon MRI was directly applied. For the local cohort, we adapted the model for an upgraded MRI acquisition using transfer learning and evaluated it in a leave-one-out process. The sCT mean absolute error for each patient was assessed. Radiotherapy dose plans based on sCT and CT were compared by assessing relevant absorbed dose differences in target volumes and organs at risk.

Results: The MAEs were 78 ± 13 HU and 76 ± 12 HU for the external and local cohort, respectively. For the external cohort, absorbed dose differences in target volumes were within ± 2.3% and within ± 1% in 95% of the cases. Differences in organs at risk were <2%. Similar results were obtained for the local cohort.

Conclusions: We have demonstrated a robust performance of a deep learning model for deriving sCT from MRI when applied to an independent external dataset. We updated the model to accommodate a larger axial field of view and software-induced changes to the input MRI. In both scenarios dose calculations based on sCT were similar to those of CT suggesting a robust and reliable method.

Original language	English
Article number	100762
Pages (from-to)	100762
Journal	Advances in Radiation Oncology
Volume	6
Issue number	6
DOIs	https://doi.org/10.1016/j.adro.2021.100762
Publication status	Published - 1 Nov 2021

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1016/j.adro.2021.100762

Fingerprint

Dive into the research topics of 'Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer'. Together they form a unique fingerprint.

Cite this

Olin, A. B., Thomas, C., Hansen, A. E., Rasmussen, J. H., Krokos, G., Urbano, T. G., Michaelidou, A., Jakoby, B., Ladefoged, C. N., Berthelsen, A. K., Håkansson, K., Vogelius, I. R., Specht, L., Barrington, S. F., Andersen, F. L., & Fischer, B. M. (2021). Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer. Advances in Radiation Oncology, 6(6), 100762. Article 100762. https://doi.org/10.1016/j.adro.2021.100762

@article{be0798020c814d56926c907d8769391b,

title = "Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer",

abstract = "Purpose: Radiotherapy planning based only on positron emission tomography/magnetic resonance imaging (PET/MRI) lacks computed tomography (CT) information required for dose calculations. In this study, a previously developed deep learning model for creating synthetic CT (sCT) from MRI in patients with head and neck cancer was evaluated in 2 scenarios: (1) using an independent external dataset, and (2) using a local dataset after an update of the model related to scanner software-induced changes to the input MRI.Methods and Materials: Six patients from an external site and 17 patients from a local cohort were analyzed separately. Each patient underwent a CT and a PET/MRI with a Dixon MRI sequence over either one (external) or 2 (local) bed positions. For the external cohort, a previously developed deep learning model for deriving sCT from Dixon MRI was directly applied. For the local cohort, we adapted the model for an upgraded MRI acquisition using transfer learning and evaluated it in a leave-one-out process. The sCT mean absolute error for each patient was assessed. Radiotherapy dose plans based on sCT and CT were compared by assessing relevant absorbed dose differences in target volumes and organs at risk.Results: The MAEs were 78 ± 13 HU and 76 ± 12 HU for the external and local cohort, respectively. For the external cohort, absorbed dose differences in target volumes were within ± 2.3% and within ± 1% in 95% of the cases. Differences in organs at risk were <2%. Similar results were obtained for the local cohort.Conclusions: We have demonstrated a robust performance of a deep learning model for deriving sCT from MRI when applied to an independent external dataset. We updated the model to accommodate a larger axial field of view and software-induced changes to the input MRI. In both scenarios dose calculations based on sCT were similar to those of CT suggesting a robust and reliable method.",

author = "Olin, {Anders B.} and Christopher Thomas and Hansen, {Adam E.} and Rasmussen, {Jacob H.} and Georgios Krokos and Urbano, {Teresa Guerrero} and Andriana Michaelidou and Bj{\"o}rn Jakoby and Ladefoged, {Claes N.} and Berthelsen, {Anne K.} and Katrin H{\aa}kansson and Vogelius, {Ivan R.} and Lena Specht and Barrington, {Sally F.} and Andersen, {Flemming L.} and Fischer, {Barbara M.}",

note = "Funding Information: Disclosures: A.B.O. reports grants from the Danish Cancer Society and grants from Siemens Healthineers, during the conduct of the study. B.J. is an employee of Siemens Healthineers. K.H. reports grants from Varian Medical Systems, outside the submitted work. I.R.V. reports grants from Varian Medical Systems, and from ViewRay Inc., outside the submitted work. L.S. reports personal fees from Takeda, personal fees from Kyowa Kirin, nonfinancial support from MSD, grants from Varian, and grants from VewRay, outside the submitted work. S.F.B. reports grants from National Institute for Health Research and Social Care, during the conduct of the study; grants from Bristol Myers Squibb international corporation, grants from Pfizer Inc, grants from Amgen Ltd, outside the submitted work. C.T., A.E.H., J.H.R., G.K., T.G.U., A.M., C.N.L., A.K.B., F.L.A., and B.M.F. have nothing to disclose. Funding Information: Sources of support: This work was partially supported by the Danish Cancer Society (R134-A8543) and Siemens Healthineers. Support is also acknowledged from the National Institute for Health Research and Social Care (NIHR) (RP-2-16-07-001). King's College London and UCL Comprehensive Cancer Imaging Centre is funded by the CRUK and EPSRC in association with the MRC and Department of Health and Social Care (England). This work was also supported by the Wellcome/EPSRC Centre for Medical Engineering at King's College London (WT 203148/Z/16/Z). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. Publisher Copyright: {\textcopyright} 2021 The Authors",

year = "2021",

month = nov,

day = "1",

doi = "10.1016/j.adro.2021.100762",

language = "English",

volume = "6",

pages = "100762",

journal = "Advances in Radiation Oncology",

publisher = "Elsevier Inc.",

number = "6",

}

Olin, AB, Thomas, C, Hansen, AE, Rasmussen, JH, Krokos, G, Urbano, TG, Michaelidou, A, Jakoby, B, Ladefoged, CN, Berthelsen, AK, Håkansson, K, Vogelius, IR, Specht, L, Barrington, SF, Andersen, FL & Fischer, BM 2021, 'Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer', Advances in Radiation Oncology, vol. 6, no. 6, 100762, pp. 100762. https://doi.org/10.1016/j.adro.2021.100762

Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer. / Olin, Anders B.; Thomas, Christopher; Hansen, Adam E. et al.
In: Advances in Radiation Oncology, Vol. 6, No. 6, 100762, 01.11.2021, p. 100762.

Research output: Contribution to journal › Article › peer-review

TY - JOUR

T1 - Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer

AU - Olin, Anders B.

AU - Thomas, Christopher

AU - Hansen, Adam E.

AU - Rasmussen, Jacob H.

AU - Krokos, Georgios

AU - Urbano, Teresa Guerrero

AU - Michaelidou, Andriana

AU - Jakoby, Björn

AU - Ladefoged, Claes N.

AU - Berthelsen, Anne K.

AU - Håkansson, Katrin

AU - Vogelius, Ivan R.

AU - Specht, Lena

AU - Barrington, Sally F.

AU - Andersen, Flemming L.

AU - Fischer, Barbara M.

N1 - Funding Information: Disclosures: A.B.O. reports grants from the Danish Cancer Society and grants from Siemens Healthineers, during the conduct of the study. B.J. is an employee of Siemens Healthineers. K.H. reports grants from Varian Medical Systems, outside the submitted work. I.R.V. reports grants from Varian Medical Systems, and from ViewRay Inc., outside the submitted work. L.S. reports personal fees from Takeda, personal fees from Kyowa Kirin, nonfinancial support from MSD, grants from Varian, and grants from VewRay, outside the submitted work. S.F.B. reports grants from National Institute for Health Research and Social Care, during the conduct of the study; grants from Bristol Myers Squibb international corporation, grants from Pfizer Inc, grants from Amgen Ltd, outside the submitted work. C.T., A.E.H., J.H.R., G.K., T.G.U., A.M., C.N.L., A.K.B., F.L.A., and B.M.F. have nothing to disclose. Funding Information: Sources of support: This work was partially supported by the Danish Cancer Society (R134-A8543) and Siemens Healthineers. Support is also acknowledged from the National Institute for Health Research and Social Care (NIHR) (RP-2-16-07-001). King's College London and UCL Comprehensive Cancer Imaging Centre is funded by the CRUK and EPSRC in association with the MRC and Department of Health and Social Care (England). This work was also supported by the Wellcome/EPSRC Centre for Medical Engineering at King's College London (WT 203148/Z/16/Z). The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care. Publisher Copyright: © 2021 The Authors

PY - 2021/11/1

Y1 - 2021/11/1

N2 - Purpose: Radiotherapy planning based only on positron emission tomography/magnetic resonance imaging (PET/MRI) lacks computed tomography (CT) information required for dose calculations. In this study, a previously developed deep learning model for creating synthetic CT (sCT) from MRI in patients with head and neck cancer was evaluated in 2 scenarios: (1) using an independent external dataset, and (2) using a local dataset after an update of the model related to scanner software-induced changes to the input MRI.Methods and Materials: Six patients from an external site and 17 patients from a local cohort were analyzed separately. Each patient underwent a CT and a PET/MRI with a Dixon MRI sequence over either one (external) or 2 (local) bed positions. For the external cohort, a previously developed deep learning model for deriving sCT from Dixon MRI was directly applied. For the local cohort, we adapted the model for an upgraded MRI acquisition using transfer learning and evaluated it in a leave-one-out process. The sCT mean absolute error for each patient was assessed. Radiotherapy dose plans based on sCT and CT were compared by assessing relevant absorbed dose differences in target volumes and organs at risk.Results: The MAEs were 78 ± 13 HU and 76 ± 12 HU for the external and local cohort, respectively. For the external cohort, absorbed dose differences in target volumes were within ± 2.3% and within ± 1% in 95% of the cases. Differences in organs at risk were <2%. Similar results were obtained for the local cohort.Conclusions: We have demonstrated a robust performance of a deep learning model for deriving sCT from MRI when applied to an independent external dataset. We updated the model to accommodate a larger axial field of view and software-induced changes to the input MRI. In both scenarios dose calculations based on sCT were similar to those of CT suggesting a robust and reliable method.

AB - Purpose: Radiotherapy planning based only on positron emission tomography/magnetic resonance imaging (PET/MRI) lacks computed tomography (CT) information required for dose calculations. In this study, a previously developed deep learning model for creating synthetic CT (sCT) from MRI in patients with head and neck cancer was evaluated in 2 scenarios: (1) using an independent external dataset, and (2) using a local dataset after an update of the model related to scanner software-induced changes to the input MRI.Methods and Materials: Six patients from an external site and 17 patients from a local cohort were analyzed separately. Each patient underwent a CT and a PET/MRI with a Dixon MRI sequence over either one (external) or 2 (local) bed positions. For the external cohort, a previously developed deep learning model for deriving sCT from Dixon MRI was directly applied. For the local cohort, we adapted the model for an upgraded MRI acquisition using transfer learning and evaluated it in a leave-one-out process. The sCT mean absolute error for each patient was assessed. Radiotherapy dose plans based on sCT and CT were compared by assessing relevant absorbed dose differences in target volumes and organs at risk.Results: The MAEs were 78 ± 13 HU and 76 ± 12 HU for the external and local cohort, respectively. For the external cohort, absorbed dose differences in target volumes were within ± 2.3% and within ± 1% in 95% of the cases. Differences in organs at risk were <2%. Similar results were obtained for the local cohort.Conclusions: We have demonstrated a robust performance of a deep learning model for deriving sCT from MRI when applied to an independent external dataset. We updated the model to accommodate a larger axial field of view and software-induced changes to the input MRI. In both scenarios dose calculations based on sCT were similar to those of CT suggesting a robust and reliable method.

UR - http://www.scopus.com/inward/record.url?scp=85122681519&partnerID=8YFLogxK

U2 - 10.1016/j.adro.2021.100762

DO - 10.1016/j.adro.2021.100762

M3 - Article

C2 - 34585026

AN - SCOPUS:85122681519

VL - 6

SP - 100762

JO - Advances in Radiation Oncology

JF - Advances in Radiation Oncology

IS - 6

M1 - 100762

ER -

Olin AB, Thomas C, Hansen AE, Rasmussen JH, Krokos G, Urbano TG et al. Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer. Advances in Radiation Oncology. 2021 Nov 1;6(6):100762. 100762. doi: 10.1016/j.adro.2021.100762

Robustness and Generalizability of Deep Learning Synthetic Computed Tomography for Positron Emission Tomography/Magnetic Resonance Imaging–Based Radiation Therapy Planning of Patients With Head and Neck Cancer

Abstract

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this