Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

MRI-only based treatment with a commercial deep-learning generation method for synthetic CT of brain

Lerner, Minna LU ; Medin, Joakim LU ; Jamtheim Gustafsson, Christian LU ; Alkner, Sara LU and Olsson, Lars E LU orcid (2020) Virtual 8th MR in RT Symposium p.47-47
Abstract
Objectives
To show feasibility of synthetic computed tomography (sCT) images generated using a commercially
available software, enabling MRI-only treatment planning for the brain in a clinical setting.
Patients and Methods
20 and 16 patients with brain malignancies, including post-surgical cases, were included for validation
and treatment, respectively. Dixon MR images of the skull were exported to the MRI Planner software
(Spectronic Medical AB), which utilizes convolutional neural network algorithms for sCT generation.
In the validation study, sCT images were rigidly registered and resampled to CT geometry for each
patient. Treatment plans were optimized on CT and retrospectively recalculated on sCT images... (More)
Objectives
To show feasibility of synthetic computed tomography (sCT) images generated using a commercially
available software, enabling MRI-only treatment planning for the brain in a clinical setting.
Patients and Methods
20 and 16 patients with brain malignancies, including post-surgical cases, were included for validation
and treatment, respectively. Dixon MR images of the skull were exported to the MRI Planner software
(Spectronic Medical AB), which utilizes convolutional neural network algorithms for sCT generation.
In the validation study, sCT images were rigidly registered and resampled to CT geometry for each
patient. Treatment plans were optimized on CT and retrospectively recalculated on sCT images for
evaluation of dosimetric and geometric endpoints. Clinical robustness in patient setup verification was
assessed by rigidly registering cone beam CT (CBCT) to sCT and CT images, respectively.
The treatment study was performed on sCT images, using CT solely for QA purposes.
Results
All sCT images were successfully generated in the validation study. Mean absolute error of the sCT
images within the body contour for all patients was 62.2 ± 4.1 HU. Average absorbed dose differences
were below 0.2%. Mean pass rate of global gamma (1%/1mm) for all patients was 100.0 ± 0.0 % within
PTV and 99.1 ± 0.6 % for the full dose distribution. No clinically relevant deviations were found in the
CBCT-sCT vs CBCT-CT image registrations. Areas of bone resection due to surgery were accurately
depicted in the sCT images. Finally, treatment success rate was 15/16. One patient was excluded due to
sCT artifacts from a haemostatic substance injected during surgery.
Conclusion
15 patients have successfully received MRI-only RT for brain tumours using the validated commercially
available sCT software. Validation showed comparable results between sCT and CT images for both
dosimetric and geometric endpoints (Less)
Please use this url to cite or link to this publication:
author
; ; ; and
organization
publishing date
type
Contribution to conference
publication status
published
subject
pages
47 - 47
conference name
Virtual 8th MR in RT Symposium
conference dates
2021-04-19 - 2021-04-21
language
English
LU publication?
yes
id
9e9ee6ef-8066-4827-beac-3e3530fe69f1
alternative location
https://www.dkfz.de/en/medphys/MRinRTHD2021/pdfs/Virtual_8th_MRinRT_AbstractBooklet_Final_20210415.pdf
date added to LUP
2022-11-15 18:00:31
date last changed
2023-05-15 09:08:45
@misc{9e9ee6ef-8066-4827-beac-3e3530fe69f1,
  abstract     = {{Objectives<br/>To show feasibility of synthetic computed tomography (sCT) images generated using a commercially<br/>available software, enabling MRI-only treatment planning for the brain in a clinical setting.<br/>Patients and Methods<br/>20 and 16 patients with brain malignancies, including post-surgical cases, were included for validation<br/>and treatment, respectively. Dixon MR images of the skull were exported to the MRI Planner software<br/>(Spectronic Medical AB), which utilizes convolutional neural network algorithms for sCT generation.<br/>In the validation study, sCT images were rigidly registered and resampled to CT geometry for each<br/>patient. Treatment plans were optimized on CT and retrospectively recalculated on sCT images for<br/>evaluation of dosimetric and geometric endpoints. Clinical robustness in patient setup verification was<br/>assessed by rigidly registering cone beam CT (CBCT) to sCT and CT images, respectively.<br/>The treatment study was performed on sCT images, using CT solely for QA purposes.<br/>Results<br/>All sCT images were successfully generated in the validation study. Mean absolute error of the sCT<br/>images within the body contour for all patients was 62.2 ± 4.1 HU. Average absorbed dose differences<br/>were below 0.2%. Mean pass rate of global gamma (1%/1mm) for all patients was 100.0 ± 0.0 % within<br/>PTV and 99.1 ± 0.6 % for the full dose distribution. No clinically relevant deviations were found in the<br/>CBCT-sCT vs CBCT-CT image registrations. Areas of bone resection due to surgery were accurately<br/>depicted in the sCT images. Finally, treatment success rate was 15/16. One patient was excluded due to<br/>sCT artifacts from a haemostatic substance injected during surgery.<br/>Conclusion<br/>15 patients have successfully received MRI-only RT for brain tumours using the validated commercially<br/>available sCT software. Validation showed comparable results between sCT and CT images for both<br/>dosimetric and geometric endpoints}},
  author       = {{Lerner, Minna and Medin, Joakim and Jamtheim Gustafsson, Christian and Alkner, Sara and Olsson, Lars E}},
  language     = {{eng}},
  pages        = {{47--47}},
  title        = {{MRI-only based treatment with a commercial deep-learning generation method for synthetic CT of brain}},
  url          = {{https://www.dkfz.de/en/medphys/MRinRTHD2021/pdfs/Virtual_8th_MRinRT_AbstractBooklet_Final_20210415.pdf}},
  year         = {{2020}},
}