Michael John James Douglass
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Michael John James Douglass has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Eva Bezak
Allied Health and Human Performance University of South Australia Adelaide South Australia Australia
13 shared publications - Ivan M. Kempson
Future Industries Institute, University of South Australia, Adelaide, Australia
7 shared publications - Alexandre M. Caraça Santos
Department of Medical Physics, Royal Adelaide Hospital, Adelaide, Australia
5 shared publications - Andrew Christopher Kennedy
School of Physical Sciences, University of Adelaide, Adelaide, South Australia 5005, Australia
4 shared publications - Annabelle M. Austin
Department of Physics University of Adelaide Adelaide SA Australia
4 shared publications - Dylan Peukert
ARC Training Ctr Integrated Operat Complex Resourc, Adelaide, SA 5005, Australia
4 shared publications
Latest Publications
- 13 June 2026
- Published by Elsevier in Physica Medica
- Vol. 148, 105835
Survey evidence demands professional guidance on large language model use in medical physics
- 13 June 2026
- Published by Elsevier in Physica Medica
- Vol. 148, 105835
Survey evidence demands professional guidance on large language model use in medical physics
- Open Access
- 24 July 2025
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 48 (4) , 1525-1538
An open-source tool for converting 3D mesh volumes into synthetic DICOM CT images for medical physics research
Access to medical imaging data is crucial for research, training, and treatment planning in medical imaging and radiation therapy. However, ethical constraints and time-consuming approval processes often limit the availability of such data for research. This study introduces DICOMator, an open-source Blender add-on designed to address this challenge by enabling the creation of synthetic CT datasets from 3D mesh objects. DICOMator aims to provide researchers and medical professionals with a flexible tool for generating customisable and semi-realistic synthetic CT data, including 4D CT datasets from user defined static or animated 3D mesh objects. The add-on leverages Blender’s powerful 3D modelling environment, utilising its mesh manipulation, animation and rendering capabilities to create synthetic data ranging from simple phantoms to accurate anatomical models. DICOMator incorporates various features to simulate common CT imaging artefacts, bridging the gap between 3D modelling and medical imaging. DICOMator voxelises 3D mesh objects, assigns appropriate Hounsfield Unit values, and applies artefact simulations. These simulations include detector noise, metal artefacts and partial volume effects. By incorporating these artefacts, DICOMator produces synthetic CT data that more closely resembles real CT scans. The resulting data is then exported in DICOM format, ensuring compatibility with existing medical imaging workflows and treatment planning systems. To demonstrate DICOMator’s capabilities, three synthetic CT datasets were created: a simple lung phantom to illustrate basic functionality, a more realistic cranial CT scan to demonstrate dose calculations and CT image registration on synthetic data in treatment planning systems. Finally, a thoracic 4D CT scan featuring multiple breathing phases was created to demonstrate the dynamic imaging capabilities and the quantitative accuracy of the synthetic datasets. These examples were chosen to highlight DICOMator’s versatility in generating diverse and complex synthetic CT data suitable for various research and educational purposes, from basic quality assurance to advanced motion management studies. DICOMator offers a promising solution to the limitations of patient CT data availability in medical physics research. By providing a user-friendly interface for creating customisable synthetic datasets from 3D meshes, it has the potential to accelerate research, validate treatment planning tools such as deformable image registration, and enhance educational resources in the field of radiation oncology medical physics. Future developments may include incorporation of other imaging modalities, such as MRI or PET, further expanding its utility in multi-modal imaging research.
- 24 July 2025
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 48 (4) , 1525-1538
Open AccessAn open-source tool for converting 3D mesh volumes into synthetic DICOM CT images for medical physics research
Access to medical imaging data is crucial for research, training, and treatment planning in medical imaging and radiation therapy. However, ethical constraints and time-consuming approval processes often limit the availability of such data for research. This study introduces DICOMator, an open-source Blender add-on designed to address this challenge by enabling the creation of synthetic CT datasets from 3D mesh objects. DICOMator aims to provide researchers and medical professionals with a flexible tool for generating customisable and semi-realistic synthetic CT data, including 4D CT datasets from user defined static or animated 3D mesh objects. The add-on leverages Blender’s powerful 3D modelling environment, utilising its mesh manipulation, animation and rendering capabilities to create synthetic data ranging from simple phantoms to accurate anatomical models. DICOMator incorporates various features to simulate common CT imaging artefacts, bridging the gap between 3D modelling and medical imaging. DICOMator voxelises 3D mesh objects, assigns appropriate Hounsfield Unit values, and applies artefact simulations. These simulations include detector noise, metal artefacts and partial volume effects. By incorporating these artefacts, DICOMator produces synthetic CT data that more closely resembles real CT scans. The resulting data is then exported in DICOM format, ensuring compatibility with existing medical imaging workflows and treatment planning systems. To demonstrate DICOMator’s capabilities, three synthetic CT datasets were created: a simple lung phantom to illustrate basic functionality, a more realistic cranial CT scan to demonstrate dose calculations and CT image registration on synthetic data in treatment planning systems. Finally, a thoracic 4D CT scan featuring multiple breathing phases was created to demonstrate the dynamic imaging capabilities and the quantitative accuracy of the synthetic datasets. These examples were chosen to highlight DICOMator’s versatility in generating diverse and complex synthetic CT data suitable for various research and educational purposes, from basic quality assurance to advanced motion management studies. DICOMator offers a promising solution to the limitations of patient CT data availability in medical physics research. By providing a user-friendly interface for creating customisable synthetic datasets from 3D meshes, it has the potential to accelerate research, validate treatment planning tools such as deformable image registration, and enhance educational resources in the field of radiation oncology medical physics. Future developments may include incorporation of other imaging modalities, such as MRI or PET, further expanding its utility in multi-modal imaging research.
- Open Access
- 10 June 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (12) , 125006
A robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties
Objective. In high-dose-rate prostate brachytherapy, uncertainties are likely to cause a deviation from the nominal treatment plan, potentially leading to failure in achieving clinical objectives. Robust optimisation has the potential to maximise the probability that objectives are met during treatment despite these uncertainties. Approach. A probabilistic robust optimiser that incorporating fourteen major uncertainty sources was developed and evaluated on 49 patients. Three objective functions were maximised to generate the approximate Pareto front of 200 robust-optimised plans, approximating the robustness of: (1) The minimum dose to the hottest 90% of the prostate ( D 90 P ), (2) The maximum dose to the urethra ( D 0.01 cc U ), and (3) The maximum dose to the rectum ( D 0 .1 cc R ). Plans were then robustly evaluated using 1000 uncertainty scenarios each simulating a possible deviation from the planned treatment. The percentage of scenarios meeting the D 90 P , D 0.01 cc U , and D 0.1 cc R metrics were determined, along with the overall pass rate, defined as the percentage of scenarios meting all three metrics simultaneously. These pass-rates, along with nominal metrics, were, were used to select the best robust-optimised plan. A radiation oncologist evaluated the best robust-optimised plans against the treatment planning system (TPS)-optimised plan for ten patients. The same selection criteria were then applied to a further cohort of 39 patients and the same plan comparisons performed. Main results. All best robust-optimised plans had higher overall pass-rates (mean: 50.7 ± 1.5%, SD: 14.2%) then TPS-optimised plans (mean: 32.0 ± 1.5%, SD: 12.3%). The mean D 0.01 cc U pass-rate was 66.0 ± 1.3% (SD: 12.1) for the robust-optimised plans compared with 47.2 ± 1.3% (SD: 9.3%) for TPS-optimised plans. The D 90 P pass-rates was higher for robust-optimised plans (mean: 85.6 ± 1.1%, SD: 9.5%) then TPS-optimised (mean: 82.2 ± 1.1%, SD: 13.8%) in 36 patients. D 0.1 cc R pass-rates remained consistently high for both optimisation methods. Significance. The robust optimisation algorithm generated plans with greater robustness than the TPS-optimised plans for nine out of ten patients evaluated by a radiation oncologist, in an average algorithm runtime of 1-minute-49 s.
- 10 June 2025
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 70 (12) , 125006
Open AccessA robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties
Objective. In high-dose-rate prostate brachytherapy, uncertainties are likely to cause a deviation from the nominal treatment plan, potentially leading to failure in achieving clinical objectives. Robust optimisation has the potential to maximise the probability that objectives are met during treatment despite these uncertainties. Approach. A probabilistic robust optimiser that incorporating fourteen major uncertainty sources was developed and evaluated on 49 patients. Three objective functions were maximised to generate the approximate Pareto front of 200 robust-optimised plans, approximating the robustness of: (1) The minimum dose to the hottest 90% of the prostate ( D 90 P ), (2) The maximum dose to the urethra ( D 0.01 cc U ), and (3) The maximum dose to the rectum ( D 0 .1 cc R ). Plans were then robustly evaluated using 1000 uncertainty scenarios each simulating a possible deviation from the planned treatment. The percentage of scenarios meeting the D 90 P , D 0.01 cc U , and D 0.1 cc R metrics were determined, along with the overall pass rate, defined as the percentage of scenarios meting all three metrics simultaneously. These pass-rates, along with nominal metrics, were, were used to select the best robust-optimised plan. A radiation oncologist evaluated the best robust-optimised plans against the treatment planning system (TPS)-optimised plan for ten patients. The same selection criteria were then applied to a further cohort of 39 patients and the same plan comparisons performed. Main results. All best robust-optimised plans had higher overall pass-rates (mean: 50.7 ± 1.5%, SD: 14.2%) then TPS-optimised plans (mean: 32.0 ± 1.5%, SD: 12.3%). The mean D 0.01 cc U pass-rate was 66.0 ± 1.3% (SD: 12.1) for the robust-optimised plans compared with 47.2 ± 1.3% (SD: 9.3%) for TPS-optimised plans. The D 90 P pass-rates was higher for robust-optimised plans (mean: 85.6 ± 1.1%, SD: 9.5%) then TPS-optimised (mean: 82.2 ± 1.1%, SD: 13.8%) in 36 patients. D 0.1 cc R pass-rates remained consistently high for both optimisation methods. Significance. The robust optimisation algorithm generated plans with greater robustness than the TPS-optimised plans for nine out of ten patients evaluated by a radiation oncologist, in an average algorithm runtime of 1-minute-49 s.
- Open Access
- 7 May 2025
- Published by Springer Nature in Research Square
An Open-Source Tool for Converting 3D Mesh volumes into Synthetic DICOM CT Images for Medical Physics Research
Background Access to medical imaging data is crucial for research, training, and treatment planning in medical imaging and radiation therapy. However, ethical constraints and time-consuming approval processes often limit the availability of such data. This study introduces DICOMator, an open-source Blender add-on designed to address this challenge by enabling the creation of synthetic CT datasets from 3D mesh objects. Purpose DICOMator aims to provide researchers and medical professionals with a flexible tool for generating customisable and qualitatively realistic synthetic CT data, including 4D CT capabilities. The add-on leverages Blender's powerful 3D modelling environment, utilising its mesh manipulation, and rendering capabilities to create synthetic data ranging from simple phantoms to accurate anatomical models. DICOMator incorporates various features to simulate common CT imaging artefacts, bridging the gap between 3D modelling and medical imaging. Methods The methodology involves voxelising 3D mesh objects, assigning appropriate Hounsfield Unit values, and applying artefact simulations. These simulations include detector noise, metal artefacts and partial volume effects. By incorporating these artefacts, DICOMator produces synthetic CT data that more closely resembles real CT scans. The resulting data is then exported in DICOM format, ensuring compatibility with existing medical imaging workflows and treatment planning systems. Results To demonstrate DICOMator's capabilities, three synthetic CT datasets were created: a simple lung phantom to illustrate basic functionality, a realistic cranial CT scan to demonstrate complex anatomical modelling, and a thoracic 4DCT scan featuring multiple breathing phases to demonstrate dynamic imaging capabilities. These examples were chosen to highlight DICOMator's versatility in generating diverse and complex synthetic CT data suitable for various research and educational purposes, from basic quality assurance to advanced motion management studies. Conclusions DICOMator offers a promising solution to the limitations of patient CT data availability in medical physics research. By providing a user-friendly interface for creating customisable synthetic datasets, it has the potential to accelerate research, improve treatment planning algorithms, and enhance educational resources in the field of radiation oncology. Future developments may include integration with other imaging modalities, such as MRI or PET, further expanding its utility in multi-modal imaging research.
- 7 May 2025
- Published by Springer Nature in Research Square
Open AccessAn Open-Source Tool for Converting 3D Mesh volumes into Synthetic DICOM CT Images for Medical Physics Research
Background Access to medical imaging data is crucial for research, training, and treatment planning in medical imaging and radiation therapy. However, ethical constraints and time-consuming approval processes often limit the availability of such data. This study introduces DICOMator, an open-source Blender add-on designed to address this challenge by enabling the creation of synthetic CT datasets from 3D mesh objects. Purpose DICOMator aims to provide researchers and medical professionals with a flexible tool for generating customisable and qualitatively realistic synthetic CT data, including 4D CT capabilities. The add-on leverages Blender's powerful 3D modelling environment, utilising its mesh manipulation, and rendering capabilities to create synthetic data ranging from simple phantoms to accurate anatomical models. DICOMator incorporates various features to simulate common CT imaging artefacts, bridging the gap between 3D modelling and medical imaging. Methods The methodology involves voxelising 3D mesh objects, assigning appropriate Hounsfield Unit values, and applying artefact simulations. These simulations include detector noise, metal artefacts and partial volume effects. By incorporating these artefacts, DICOMator produces synthetic CT data that more closely resembles real CT scans. The resulting data is then exported in DICOM format, ensuring compatibility with existing medical imaging workflows and treatment planning systems. Results To demonstrate DICOMator's capabilities, three synthetic CT datasets were created: a simple lung phantom to illustrate basic functionality, a realistic cranial CT scan to demonstrate complex anatomical modelling, and a thoracic 4DCT scan featuring multiple breathing phases to demonstrate dynamic imaging capabilities. These examples were chosen to highlight DICOMator's versatility in generating diverse and complex synthetic CT data suitable for various research and educational purposes, from basic quality assurance to advanced motion management studies. Conclusions DICOMator offers a promising solution to the limitations of patient CT data availability in medical physics research. By providing a user-friendly interface for creating customisable synthetic datasets, it has the potential to accelerate research, improve treatment planning algorithms, and enhance educational resources in the field of radiation oncology. Future developments may include integration with other imaging modalities, such as MRI or PET, further expanding its utility in multi-modal imaging research.
- Open Access
- 23 February 2024
- Published by Springer Nature in Research Square
A robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties II: Algorithm evaluation on 49 patients
Introduction: Treatment plans in high-dose-rate (HDR) prostate brachytherapy (BT) are currently optimised to meet clinical objectives in the nominal plan; however, uncertainties cause the scenario delivered to deviate from this plan and may fail objectives. A robust optimiser generates plans that maximise the probability of dose-volume-histogram (DVH) metrics passing in the delivered scenario. Method: A probabilistic robust optimiser was evaluated on 49 patients. Three objective functions were maximised to obtain the Pareto front consisting of 200 robust-optimised treatment plans. Each objective function approximated the robustness of three DVH metrics: (1) the minimum dose to the hottest 90% of the prostate, \({\text{D}}_{90}^{\text{P}}\), the maximum doses to the (2) urethra, \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\), and (3) rectum, \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\). Pareto front plans were robustly evaluated using 1000 probabilistic uncertainty scenarios, outputting the pass-rates for \({\text{D}}_{90}^{\text{P}}\), \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\), \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\), and all three simultaneously, the overall pass-rate. The best robust-optimised plan was selected from the Pareto front and compared to the TPS-optimised plan for ten patients by a radiation oncologist, then 49 patient robust-optimised plans were compared. Results: All robust-optimised plans had higher overall pass-rates (mean: 50.7 ± 1.5%, SD: 14.2%) compared to the TPS-optimised plans (mean: 32.0 ± 1.5%, SD: 12.3%). This was also the case with the \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\) pass-rates with the robust-optimised plans having a mean of 66.0 ± 1.3% (SD: 12.1) compared with 47.2 ± 1.3% (SD: 9.3%). The pass-rates for \({\text{D}}_{90}^{\text{P}}\) were higher for the robust-optimised plans (mean: 85.6 ± 1.1%, SD: 9.5%) in 36 patients, the TPS-optimised mean was 82.2 ± 1.1% (SD: 13.8%). Twenty-seven patients had higher \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\) pass-rates in the robust-optimised plans with a mean of 94.54 ± 0.56% (SD: 7.35%) compared with 93.71 ± 0.56% (SD: 6.73%). Conclusion: A robust optimisation algorithm generated treatment plans with higher robustness than the TPS-optimised plans for nine out of ten patients, evaluated by a radiation oncologist, in an average time of 1-minute-49-seconds.- 23 February 2024
- Published by Springer Nature in Research Square
Open AccessA robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties II: Algorithm evaluation on 49 patients
Introduction: Treatment plans in high-dose-rate (HDR) prostate brachytherapy (BT) are currently optimised to meet clinical objectives in the nominal plan; however, uncertainties cause the scenario delivered to deviate from this plan and may fail objectives. A robust optimiser generates plans that maximise the probability of dose-volume-histogram (DVH) metrics passing in the delivered scenario. Method: A probabilistic robust optimiser was evaluated on 49 patients. Three objective functions were maximised to obtain the Pareto front consisting of 200 robust-optimised treatment plans. Each objective function approximated the robustness of three DVH metrics: (1) the minimum dose to the hottest 90% of the prostate, \({\text{D}}_{90}^{\text{P}}\), the maximum doses to the (2) urethra, \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\), and (3) rectum, \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\). Pareto front plans were robustly evaluated using 1000 probabilistic uncertainty scenarios, outputting the pass-rates for \({\text{D}}_{90}^{\text{P}}\), \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\), \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\), and all three simultaneously, the overall pass-rate. The best robust-optimised plan was selected from the Pareto front and compared to the TPS-optimised plan for ten patients by a radiation oncologist, then 49 patient robust-optimised plans were compared. Results: All robust-optimised plans had higher overall pass-rates (mean: 50.7 ± 1.5%, SD: 14.2%) compared to the TPS-optimised plans (mean: 32.0 ± 1.5%, SD: 12.3%). This was also the case with the \({\text{D}}_{0.01 \text{c}\text{c}}^{\text{U}}\) pass-rates with the robust-optimised plans having a mean of 66.0 ± 1.3% (SD: 12.1) compared with 47.2 ± 1.3% (SD: 9.3%). The pass-rates for \({\text{D}}_{90}^{\text{P}}\) were higher for the robust-optimised plans (mean: 85.6 ± 1.1%, SD: 9.5%) in 36 patients, the TPS-optimised mean was 82.2 ± 1.1% (SD: 13.8%). Twenty-seven patients had higher \({\text{D}}_{0.1 \text{c}\text{c}}^{\text{R}}\) pass-rates in the robust-optimised plans with a mean of 94.54 ± 0.56% (SD: 7.35%) compared with 93.71 ± 0.56% (SD: 6.73%). Conclusion: A robust optimisation algorithm generated treatment plans with higher robustness than the TPS-optimised plans for nine out of ten patients, evaluated by a radiation oncologist, in an average time of 1-minute-49-seconds. - Open Access
- 20 February 2024
- Published by Springer Nature in Research Square
A robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties I: Algorithm implementation
Introduction In high-dose-rate prostate brachytherapy, uncertainties cause a deviation from the nominal treatment plan, leading to a possible failure of clinical objectives in the delivered scenario. Robust optimisation has the potential to maximise the probability that these objectives are met during treatment. Method A computationally efficient probabilistic robust optimisation algorithm was developed and evaluated comprehensively on one patient by comparing it to the treatment-planning-systems (TPS) optimised plan. Three objective functions were maximised within a genetic algorithm (NSGA-ii), each an approximation for robustness against uncertainty for three clinical objectives: the minimum dose to the hottest 90% of the prostate target, , and the maximum doses to the urethra, ,and the rectum, . The approximations are derived from a probabilistic robust evaluation algorithm incorporating 14 major planning and treatment uncertainties. The robustness of a plan was quantified as a pass-rate from 500 probabilistic uncertainty scenarios for , and .Two hundred robust-optimised plans are generated that are the best trade-off between the three-competing DVH metric pass-rates. Results The robust-optimised plans on average (mean) increased in overall robustness by 58.5±3.0%(SD: 7.1%, min: 34.1%, max: 67.7%) compared to the TPS-optimised plan. The robustness increase for the pass-rate was 31.2±2.2%(SD: 15.6%, min: -5.1%, max: 46.7%), for the pass-rate, the increase was 48.2±2.6%(SD: 11.9%, min: 26.9%, max: 67.7%), and for the pass-rate, the change was 0.0±1.1%(SD: 0.72%, min: -2.6%, max: 0.4%). Conclusion The robust optimisation algorithm was demonstrated to produce more robust plans than the TPS, in an increased probability of target coverage and organs-at-risk sparing within a clinically reasonable time.
- 20 February 2024
- Published by Springer Nature in Research Square
Open AccessA robust optimisation genetic algorithm for HDR prostate brachytherapy including all major uncertainties I: Algorithm implementation
Introduction In high-dose-rate prostate brachytherapy, uncertainties cause a deviation from the nominal treatment plan, leading to a possible failure of clinical objectives in the delivered scenario. Robust optimisation has the potential to maximise the probability that these objectives are met during treatment. Method A computationally efficient probabilistic robust optimisation algorithm was developed and evaluated comprehensively on one patient by comparing it to the treatment-planning-systems (TPS) optimised plan. Three objective functions were maximised within a genetic algorithm (NSGA-ii), each an approximation for robustness against uncertainty for three clinical objectives: the minimum dose to the hottest 90% of the prostate target, , and the maximum doses to the urethra, ,and the rectum, . The approximations are derived from a probabilistic robust evaluation algorithm incorporating 14 major planning and treatment uncertainties. The robustness of a plan was quantified as a pass-rate from 500 probabilistic uncertainty scenarios for , and .Two hundred robust-optimised plans are generated that are the best trade-off between the three-competing DVH metric pass-rates. Results The robust-optimised plans on average (mean) increased in overall robustness by 58.5±3.0%(SD: 7.1%, min: 34.1%, max: 67.7%) compared to the TPS-optimised plan. The robustness increase for the pass-rate was 31.2±2.2%(SD: 15.6%, min: -5.1%, max: 46.7%), for the pass-rate, the increase was 48.2±2.6%(SD: 11.9%, min: 26.9%, max: 67.7%), and for the pass-rate, the change was 0.0±1.1%(SD: 0.72%, min: -2.6%, max: 0.4%). Conclusion The robust optimisation algorithm was demonstrated to produce more robust plans than the TPS, in an increased probability of target coverage and organs-at-risk sparing within a clinically reasonable time.
- Open Access
- 30 May 2023
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 46 (3) , 1115-1130
Being certain about uncertainties: a robust evaluation method for high-dose-rate prostate brachytherapy treatment plans including the combination of uncertainties
In high-dose-rate (HDR) prostate brachytherapy the combined effect of uncertainties cause a range of possible dose distributions deviating from the nominal plan, and which are not considered during treatment plan evaluation. This could lead to dosimetric misses for critical structures and overdosing of organs at risk. A robust evaluation method to assess the combination of uncertainties during plan evaluation is presented and demonstrated on one HDR prostate ultrasound treatment plan retrospectively. A range of uncertainty scenarios are simulated by changing six parameters in the nominal plan and calculating the corresponding dose distribution. Two methods are employed to change the parameters, a probabilistic approach using random number sampling to evaluate the likelihood of variation in dose distributions, and a combination of the most extreme possible values to access the worst-case dosimetric outcomes. One thousand probabilistic scenarios were run on the single treatment plan with 43.2% of scenarios passing seven of the eight clinical objectives. The prostate D90 had a standard deviation of 4.4%, with the worst case decreasing the dose by up to 27.2%. The urethra D10 was up to 29.3% higher than planned in the worst case. All DVH metrics in the probabilistic scenarios were found to be within acceptable clinical constraints for the plan under statistical tests for significance. The clinical significance of the results from the robust evaluation method presented on any individual treatment plan needs to be compared in the context of a historical data set that contains patient outcomes with robustness analysis data to ascertain a baseline acceptance.
- 30 May 2023
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 46 (3) , 1115-1130
Open AccessBeing certain about uncertainties: a robust evaluation method for high-dose-rate prostate brachytherapy treatment plans including the combination of uncertainties
In high-dose-rate (HDR) prostate brachytherapy the combined effect of uncertainties cause a range of possible dose distributions deviating from the nominal plan, and which are not considered during treatment plan evaluation. This could lead to dosimetric misses for critical structures and overdosing of organs at risk. A robust evaluation method to assess the combination of uncertainties during plan evaluation is presented and demonstrated on one HDR prostate ultrasound treatment plan retrospectively. A range of uncertainty scenarios are simulated by changing six parameters in the nominal plan and calculating the corresponding dose distribution. Two methods are employed to change the parameters, a probabilistic approach using random number sampling to evaluate the likelihood of variation in dose distributions, and a combination of the most extreme possible values to access the worst-case dosimetric outcomes. One thousand probabilistic scenarios were run on the single treatment plan with 43.2% of scenarios passing seven of the eight clinical objectives. The prostate D90 had a standard deviation of 4.4%, with the worst case decreasing the dose by up to 27.2%. The urethra D10 was up to 29.3% higher than planned in the worst case. All DVH metrics in the probabilistic scenarios were found to be within acceptable clinical constraints for the plan under statistical tests for significance. The clinical significance of the results from the robust evaluation method presented on any individual treatment plan needs to be compared in the context of a historical data set that contains patient outcomes with robustness analysis data to ascertain a baseline acceptance.
- Open Access
- 8 February 2023
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 46 (1) , 367-375
Synthetic cranial MRI from 3D optical surface scans using deep learning for radiation therapy treatment planning
Background: Optical scanning technologies are increasingly being utilised to supplement treatment workflows in radiation oncology, such as surface-guided radiotherapy or 3D printing custom bolus. One limitation of optical scanning devices is the absence of internal anatomical information of the patient being scanned. As a result, conventional radiation therapy treatment planning using this imaging modality is not feasible. Deep learning is useful for automating various manual tasks in radiation oncology, most notably, organ segmentation and treatment planning. Deep learning models have also been used to transform MRI datasets into synthetic CT datasets, facilitating the development of MRI-only radiation therapy planning. Aims: To train a pix2pix generative adversarial network to transform 3D optical scan data into estimated MRI datasets for a given patient to provide additional anatomical data for a select few radiation therapy treatment sites. The proposed network may provide useful anatomical information for treatment planning of surface mould brachytherapy, total body irradiation, and total skin electron therapy, for example, without delivering any imaging dose. Methods: A 2D pix2pix GAN was trained on 15,000 axial MRI slices of healthy adult brains paired with corresponding external mask slices. The model was validated on a further 5000 previously unseen external mask slices. The predictions were compared with the “ground-truth” MRI slices using the multi-scale structural similarity index (MSSI) metric. A certified neuro-radiologist was subsequently consulted to provide an independent review of the model’s performance in terms of anatomical accuracy and consistency. The network was then applied to a 3D photogrammetry scan of a test subject to demonstrate the feasibility of this novel technique. Results: The trained pix2pix network predicted MRI slices with a mean MSSI of 0.831 ± 0.057 for the 5000 validation images indicating that it is possible to estimate a significant proportion of a patient’s gross cranial anatomy from a patient’s exterior contour. When independently reviewed by a certified neuro-radiologist, the model’s performance was described as “quite amazing, but there are limitations in the regions where there is wide variation within the normal population.” When the trained network was applied to a 3D model of a human subject acquired using optical photogrammetry, the network could estimate the corresponding MRI volume for that subject with good qualitative accuracy. However, a ground-truth MRI baseline was not available for quantitative comparison. Conclusions: A deep learning model was developed, to transform 3D optical scan data of a patient into an estimated MRI volume, potentially increasing the usefulness of optical scanning in radiation therapy planning. This work has demonstrated that much of the human cranial anatomy can be predicted from the external shape of the head and may provide an additional source of valuable imaging data. Further research is required to investigate the feasibility of this approach for use in a clinical setting and further improve the model’s accuracy.
- 8 February 2023
- Published by Springer Nature in Physical and Engineering Sciences in Medicine
- Vol. 46 (1) , 367-375
Open AccessSynthetic cranial MRI from 3D optical surface scans using deep learning for radiation therapy treatment planning
Background: Optical scanning technologies are increasingly being utilised to supplement treatment workflows in radiation oncology, such as surface-guided radiotherapy or 3D printing custom bolus. One limitation of optical scanning devices is the absence of internal anatomical information of the patient being scanned. As a result, conventional radiation therapy treatment planning using this imaging modality is not feasible. Deep learning is useful for automating various manual tasks in radiation oncology, most notably, organ segmentation and treatment planning. Deep learning models have also been used to transform MRI datasets into synthetic CT datasets, facilitating the development of MRI-only radiation therapy planning. Aims: To train a pix2pix generative adversarial network to transform 3D optical scan data into estimated MRI datasets for a given patient to provide additional anatomical data for a select few radiation therapy treatment sites. The proposed network may provide useful anatomical information for treatment planning of surface mould brachytherapy, total body irradiation, and total skin electron therapy, for example, without delivering any imaging dose. Methods: A 2D pix2pix GAN was trained on 15,000 axial MRI slices of healthy adult brains paired with corresponding external mask slices. The model was validated on a further 5000 previously unseen external mask slices. The predictions were compared with the “ground-truth” MRI slices using the multi-scale structural similarity index (MSSI) metric. A certified neuro-radiologist was subsequently consulted to provide an independent review of the model’s performance in terms of anatomical accuracy and consistency. The network was then applied to a 3D photogrammetry scan of a test subject to demonstrate the feasibility of this novel technique. Results: The trained pix2pix network predicted MRI slices with a mean MSSI of 0.831 ± 0.057 for the 5000 validation images indicating that it is possible to estimate a significant proportion of a patient’s gross cranial anatomy from a patient’s exterior contour. When independently reviewed by a certified neuro-radiologist, the model’s performance was described as “quite amazing, but there are limitations in the regions where there is wide variation within the normal population.” When the trained network was applied to a 3D model of a human subject acquired using optical photogrammetry, the network could estimate the corresponding MRI volume for that subject with good qualitative accuracy. However, a ground-truth MRI baseline was not available for quantitative comparison. Conclusions: A deep learning model was developed, to transform 3D optical scan data of a patient into an estimated MRI volume, potentially increasing the usefulness of optical scanning in radiation therapy planning. This work has demonstrated that much of the human cranial anatomy can be predicted from the external shape of the head and may provide an additional source of valuable imaging data. Further research is required to investigate the feasibility of this approach for use in a clinical setting and further improve the model’s accuracy.
- Open Access
- 13 September 2022
- Published by Springer Nature in Research Square
Predicting cranial MRI anatomy from 3D optical surface scans using deep learning for radiation therapy treatment planning
Background Optical scanning technologies are increasingly being utilised to supplement treatment workflows in radiation oncology, such as surface-guided radiotherapy or 3D printing custom bolus. One limitation of optical scanning devices is the absence of internal anatomical information of the patient being scanned. As a result, conventional radiation therapy treatment planning using this imaging modality is not feasible. Deep learning is useful for automating various manual tasks in radiation oncology, most notably, organ segmentation and treatment planning. Deep learning models have also been used to transform MRI datasets into synthetic CT datasets, facilitating the development of MRI-only radiation therapy planning. Aims To train a pix2pix generative adversarial network was trained to transform 3D optical scan data into estimated MRI datasets for a given patient to provide additional anatomical data for a select few radiation therapy treatment sites. The proposed network may provide useful anatomical information for treatment planning of surface mould brachytherapy, total body irradiation, and total skin electron therapy, for example, without delivering any imaging dose. Methods A 2D pix2pix GAN was trained on 15,000 axial MRI slices of healthy adult brains paired with corresponding external mask slices. The model was validated on a further 5000 previously unseen external mask slices. The predictions were compared with the “ground-truth” MRI slices using the multi-scale structural similarity index (MSSI) metric. A certified neuro-radiologist was subsequently consulted to provide an independent review of the model’s performance in terms of anatomical accuracy and consistency. The network was then applied to a 3D photogrammetry scan of a test subject to demonstrate the feasibility of this novel technique. Results The trained pix2pix network predicted MRI slices with a mean MSSI of 0.831 ± 0.057 for the 5000 validation images indicating that it is possible to estimate a significant proportion of a patient’s gross cranial anatomy from a patient’s exterior contour. When independently reviewed by a certified neuro-radiologist, the model’s performance was described as “quite amazing, but there are limitations in the regions where there is wide variation within the normal population.” When the trained network was applied to a 3D model of a human subject acquired using optical photogrammetry, the network could estimate the corresponding MRI volume for that subject with good qualitative accuracy. However, a ground-truth MRI baseline was not available for quantitative comparison. Conclusions A deep learning model was developed, to transform 3D optical scan data of a patient into an estimated MRI volume, potentially increasing the usefulness of optical scanning in radiation therapy planning. This work has demonstrated that much of the human cranial anatomy can be predicted from the external shape of the head and may provide an additional source of valuable imaging data. Further research is required to investigate the feasibility of this approach for use in a clinical setting and further improve the model’s accuracy.
- 13 September 2022
- Published by Springer Nature in Research Square
Open AccessPredicting cranial MRI anatomy from 3D optical surface scans using deep learning for radiation therapy treatment planning
Background Optical scanning technologies are increasingly being utilised to supplement treatment workflows in radiation oncology, such as surface-guided radiotherapy or 3D printing custom bolus. One limitation of optical scanning devices is the absence of internal anatomical information of the patient being scanned. As a result, conventional radiation therapy treatment planning using this imaging modality is not feasible. Deep learning is useful for automating various manual tasks in radiation oncology, most notably, organ segmentation and treatment planning. Deep learning models have also been used to transform MRI datasets into synthetic CT datasets, facilitating the development of MRI-only radiation therapy planning. Aims To train a pix2pix generative adversarial network was trained to transform 3D optical scan data into estimated MRI datasets for a given patient to provide additional anatomical data for a select few radiation therapy treatment sites. The proposed network may provide useful anatomical information for treatment planning of surface mould brachytherapy, total body irradiation, and total skin electron therapy, for example, without delivering any imaging dose. Methods A 2D pix2pix GAN was trained on 15,000 axial MRI slices of healthy adult brains paired with corresponding external mask slices. The model was validated on a further 5000 previously unseen external mask slices. The predictions were compared with the “ground-truth” MRI slices using the multi-scale structural similarity index (MSSI) metric. A certified neuro-radiologist was subsequently consulted to provide an independent review of the model’s performance in terms of anatomical accuracy and consistency. The network was then applied to a 3D photogrammetry scan of a test subject to demonstrate the feasibility of this novel technique. Results The trained pix2pix network predicted MRI slices with a mean MSSI of 0.831 ± 0.057 for the 5000 validation images indicating that it is possible to estimate a significant proportion of a patient’s gross cranial anatomy from a patient’s exterior contour. When independently reviewed by a certified neuro-radiologist, the model’s performance was described as “quite amazing, but there are limitations in the regions where there is wide variation within the normal population.” When the trained network was applied to a 3D model of a human subject acquired using optical photogrammetry, the network could estimate the corresponding MRI volume for that subject with good qualitative accuracy. However, a ground-truth MRI baseline was not available for quantitative comparison. Conclusions A deep learning model was developed, to transform 3D optical scan data of a patient into an estimated MRI volume, potentially increasing the usefulness of optical scanning in radiation therapy planning. This work has demonstrated that much of the human cranial anatomy can be predicted from the external shape of the head and may provide an additional source of valuable imaging data. Further research is required to investigate the feasibility of this approach for use in a clinical setting and further improve the model’s accuracy.
- Open Access
- 2 April 2022
- Published by Wiley in Journal of Medical Radiation Sciences
- Vol. 69 (2) , 139-142
Can optical scanning technologies replace CT for 3D printed medical devices in radiation oncology?
3D printing is being increasingly adopted in radiation oncology for printing highly conformal medical devices for treatment. Optical surface reconstruction technologies have been shown to be useful for 3D printing applications due to their higher spatial resolution, non‐ionising radiation imaging and will likely supplement existing radiographic imaging techniques in the future.
- 2 April 2022
- Published by Wiley in Journal of Medical Radiation Sciences
- Vol. 69 (2) , 139-142
Open AccessCan optical scanning technologies replace CT for 3D printed medical devices in radiation oncology?
3D printing is being increasingly adopted in radiation oncology for printing highly conformal medical devices for treatment. Optical surface reconstruction technologies have been shown to be useful for 3D printing applications due to their higher spatial resolution, non‐ionising radiation imaging and will likely supplement existing radiographic imaging techniques in the future.