Добавил:
Sekretar
kiopkiopkiop18@yandex.ru
t.me/Prokururor I Вовсе не секретарь, но почту проверяю
Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз:
Предмет:
Файл:Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5525_Библиотеки_им_академика_М_И_Перельмана.pdf
X
- •Foreword
- •Acknowledgments
- •Editor biographies
- •Yi Wang
- •X. Sharon Qi
- •List of contributors
- •1.2.3 Feature engineering and representation
- •1.2.4 Linear separability
- •1.2.5 Classical models
- •1.1 A brief introduction to AI
- •1.2 Machine learning basics
- •1.2.1 Learning paradigms
- •1.3 Artificial neural networks
- •1.3.1 Feed-forward neural networks
- •1.3.2 Recurrent neural networks
- •1.3.3 Convolutional neural networks
- •1.3.4 Attention
- •1.3.5 Training neural networks
- •1.3.6 Applications and use cases of deep learning
- •1.4 Model training and evaluation
- •1.4.1 Hyperparameters
- •1.4.2 Data split
- •1.4.3 Evaluation metrics
- •1.5 Generative models
- •1.5.1 Generative adversarial networks
- •1.5.2 Diffusion models
- •1.5.3 Applications and use cases
- •1.6 Ethical consideration and bias
- •1.6.1 Transparency and explainability
- •1.6.2 Bias and fairness
- •1.6.3 Data privacy violation
- •1.6.4 Risk and misuse
- •1.7 Summary
- •References
- •2.1 Introduction
- •2.1.2 Staff roles in radiation therapy
- •2.2 Overview of AI in radiation therapy
- •2.2.1 Patient evaluation and dose prescription
- •2.2.2 Treatment simulation
- •2.2.3 Contouring
- •2.2.4 Treatment planning
- •2.2.5 Quality assurance
- •2.2.6 Treatment delivery
- •2.2.7 Response assessment and toxicity management
- •2.3 Summary
- •3.1 Introduction
- •3.1.1 Introduction of clinical decision making and AI
- •3.1.2 The role of AI in clinical decision making
- •3.2 AI algorithms for clinical decision making
- •3.2.2 Radiomics
- •3.2.3 Data integration by AI
- •3.2.4 Interpretability of AI models
- •3.3 Application of AI in clinical decision making
- •3.3.1 Diagnosis and disease phenotyping
- •3.3.2 Personalized treatment
- •3.3.3 Treatment outcome and prognosis prediction
- •3.4 Challenges and future directions of AI in clinical decision making
- •3.4.1 Challenges and concerns
- •3.4.2 Future directions
- •3.5 Summary
- •References
- •4.1 Introduction
- •4.2 Imaging for treatment planning
- •4.2.1 CT simulation
- •4.2.2 4D-CT
- •4.2.3 PET/CT
- •4.2.4 MRI
- •4.3 Imaging for treatment guidance
- •4.3.1 Portal imaging
- •4.3.2 CBCT
- •4.3.3 CT-on-rail and CT-linac
- •4.3.4 MR-linac
- •4.3.5 PET-linac
- •4.4 Imaging for motion management
- •4.4.1 ExacTrac
- •4.4.2 Varian triggered imaging
- •4.4.3 4D-CBCT
- •4.4.4 Cine MRI
- •4.4.5 4D-MRI
- •4.4.6 Surface imaging
- •4.5 Imaging for treatment assessment
- •4.5.1 Contrasted CT
- •4.5.2 PET/CT
- •4.5.3 Functional MRI
- •4.6 Summary
- •5.1 Introduction to big data in radiation oncology
- •5.1.1 Overview of big data
- •5.1.2 Sources of big data in radiation oncology
- •5.1.3 Big data and AI in radiation oncology
- •5.2 Big data lifecycle in radiation oncology
- •5.2.1 Data aggregation and storage
- •5.2.2 Data sharing and security
- •5.4 The application of big data in radiation oncology
- •5.4.1 Medical image segmentation
- •5.4.2 Automatic treatment planning
- •5.4.3 Treatment response prediction
- •5.4.4 Quality assurance and patient safety
- •5.4.5 Clinical decision support
- •5.2.3 Data visualization
- •5.2.4 Knowledge creation and implementation
- •5.2.5 Data archiving and deletion
- •5.3 Big data analytics with AI
- •5.3.1 Data processing and integration
- •5.3.2 AI modeling
- •5.5 Challenges and future perspectives
- •5.6 Summary
- •Reference
- •6.1 The road to ART
- •6.1.1 3D conformal radiotherapy (3DCRT)
- •6.1.2 Intensity modulated radiotherapy (IMRT)
- •6.1.3 Image-guided radiotherapy (IGRT)
- •6.1.4 Adaptive radiotherapy (ART)
- •6.2 ART workflow and implementation
- •6.2.2 Current practice
- •6.2.3 Clinical impact
- •6.3 Considerations for implementing online ART
- •6.3.1 Time as a limiting factor
- •6.3.2 Implications for fast and reliable re-planning
- •6.3.4 Clinical considerations
- •6.4 Summary
- •7.1 Components of ART workflow
- •7.1.1 Simulation
- •7.1.2 Pre-planning
- •7.1.3 Online imaging and daily re-planning
- •7.1.4 Quality assurance
- •7.2 AI-driven ART
- •7.2.1 Simulation
- •7.2.2 Pre-planning
- •7.2.3 AI for delivery
- •7.3 Outlook and future directions
- •7.3.1 Real-time ART with AI
- •7.3.2 Dose escalation and functional adaption with AI
- •7.4 Summary
- •References
- •8.1 Introduction
- •8.2 Synthetic CT: deep learning methods
- •8.2.1 Conventional methods
- •8.2.2 U-Net
- •8.2.3 Generative adversarial networks
- •8.2.4 Denoising diffusion probabilistic model
- •8.3 Synthetic CT from CBCT
- •8.3.1 Noise and artifact reduction
- •8.3.2 Online dose calculation
- •8.3.3 Online image segmentation
- •8.4 Synthetic CT from MRI
- •8.4.1 Synthetic image accuracy
- •8.4.2 Dose calculation in MR-only radiation therapy
- •8.4.3 PET attenuation correction
- •8.4.4 Image registration
- •8.5 Discussion and outlook
- •8.6 Summary
- •References
- •9.1 AI-based image registration and segmentation for ART
- •9.1.1 Adaptive radiation therapy
- •9.2 Artificial intelligence
- •9.2.1 What is machine learning?
- •9.2.2 What is deep learning?
- •9.3 Deep learning: the basic components
- •9.3.1 Convolutional neural networks: looking at the picture
- •9.3.2 Pooling layers: keeping what matters most
- •9.3.3 Fully connected (dense) layers: bringing it all together
- •9.3.4 Activations
- •9.3.5 Loss: driving the model
- •9.3.6 Auto-encoders: remove the noise
- •9.3.7 Supervised versus unsupervised learning
- •9.3.8 Pre-trained convolutional neural networks
- •9.4 Image registration: bringing two images together
- •9.4.1 Registration similarity metrics
- •9.4.2 Types of registrations
- •9.5 AI-based image registration
- •9.5.1 Supervised learning
- •9.5.2 Unsupervised learning
- •9.5.3 Registration in ART
- •9.5.4 Commonalities in architectures
- •9.6 Image segmentation
- •9.6.1 Introduction: coloring by the numbers
- •9.6.2 Segmentation networks
- •9.6.3 Best practices
- •9.7 Summary
- •10.1 Introduction
- •10.1.1 Overview of chapter content
- •10.2 The landscape of AI-assisted dose prediction
- •10.2.1 Traditional machine learning for dose prediction
- •10.2.2 Deep learning-based dose prediction
- •10.2.3 Challenges in AI-assisted dose prediction
- •10.3 Re-planning workflows powered by AI
- •10.3.1 Deep learning for re-planning pipelines
- •10.4 Future directions of AI-assisted dose prediction and re-planning
- •10.5 Summary
- •11.1 Introduction
- •11.2.1 Imaging-based motion monitoring
- •11.2.2 Delivery system actions
- •11.2.3 Challenges for real-time ART implementation
- •11.3 AI in real-time ART workflows
- •11.3.1 Improving intrafraction motion monitoring through AI
- •11.3.2 Mitigating system latency through AI
- •11.4 AI for ART delivery: future directions
- •11.4.1 Management of non-respiratory motion
- •11.4.2 Training AI models with small or unpaired datasets
- •11.4.4 Biology-guided ART delivery
- •11.5 Summary
- •References
- •12.1 Introduction
- •12.2 Patient QA
- •12.2.1 Pre-planning QA
- •12.2.2 Pre-treatment plan QA
- •12.2.3 On-treatment QA
- •12.3 Treatment delivery systems and instruments
- •12.3.1 Machine commissioning
- •12.3.2 Machine QA
- •12.3.3 Dosimetry tool QA
- •12.4 Summary
- •References
- •13.1 Data resources for response modeling in radiotherapy
- •13.1.1 Clinical data
- •13.1.2 Imaging (radiomics)
- •13.1.3 Treatment planning (dosiomics)
- •13.1.4 Multiomics
- •13.2 Radiotherapy treatment outcome modeling
- •13.2.1 TCP/NTCP in radiotherapy
- •13.2.2 Clinical outcomes versus PROs
- •13.2.3 Machine learning response prediction
- •13.2.4 Explainability of ML response models
- •13.2.5 Sample use cases
- •13.3 AI response-based adaptive radiotherapy
- •13.3.1 Requirements and challenges
- •13.3.2 Prediction versus treatment optimization
- •13.3.3 Sample use cases
- •13.4 Challenges and recommendations
- •13.5 Summary
- •Acknowledgments
- •References
- •14.1 Overview of challenges in AI-driven ART
- •14.2 Data challenges
- •14.2.1 Data availability
- •14.2.2 Data quality
- •14.2.3 Data privacy
- •14.3 Technical challenges
- •14.3.2 Model robustness and generalizability
- •14.3.3 Model explainability and interpretability
- •14.4 Challenges associated with online and real-time workflows
- •14.4.1 Image quality
- •14.4.2 Dose calculation
- •14.4.3 Real-time ART
- •14.5 Operational challenges
- •14.5.1 Clinical validation
- •14.5.3 Staff training
- •14.5.4 User experiences
- •14.5.5 Quality management program
- •14.5.6 Financial challenges
- •14.6 Ethical, regulatory, and legal challenges
- •14.6.1 Ethical issues
- •14.6.2 Regulatory and legal issues
- •14.7 Summary
- •References
- •15.1 Clinical considerations for CT-based offline ART
- •15.1.1 Patient and site selection
- •15.1.2 Re-simulation
- •15.1.3 Re-planning
- •15.1.4 Plan summation and evaluation
- •15.1.6 Limitations and future directions
- •15.2 Clinical considerations for CBCT/CT-based online ART
- •15.2.2 Patient and site selection
- •15.2.3 Simulation
- •15.2.4 Pre-planning review
- •15.2.5 Reference planning
- •15.2.9 Limitations and future directions
- •15.3 Summary
- •References
- •16.1 Introduction
- •16.2 Overview of MRI-guided ART systems
- •16.3 MRI-guided ART workflow
- •16.4 AI applications for MRI-guided ART
- •16.4.1 Synthetic CT generation
- •References
- •16.4.2 Auto-segmentation
- •16.4.3 Image registration
- •16.4.4 Others
- •16.4.5 Future AI development and implementation
- •16.5 Summary
- •17.1 Functional PET-guided ART
- •17.1.1 PET-based functional imaging overview
- •17.1.2 From anatomy to function: the power of PET in radiation therapy
- •17.1.5 Conclusions and future prospects
- •17.2 Functional MRI-guided ART
- •17.2.1 From anatomy to function: the power of functional MRI in radiation therapy
- •17.2.4 Conclusion and future prospects
- •17.3 Summary
- •References
- •18.1 Proton ART
- •18.1.1 Clinical context and necessity
- •18.1.2 Patient populations
- •18.1.4 Rationale for AI in proton ART
- •18.2 AI in proton ART
- •18.2.1 Imaging
- •18.2.2 Deformable and rigid registration
- •18.2.3 Contour propagation
- •18.2.4 Dose calculations
- •18.2.5 Plan optimization
- •18.2.6 Other developments
- •18.3 Implementation of adaptive proton therapy
- •18.4 Summary
- •References
- •19.1 Designing clinical trials with AI
- •19.1.1 The essential role of clinical trials
- •19.1.2 Trial protocols and methodologies
- •19.1.3 AI-driven clinical trial design and execution
- •19.1.4 Incorporation of digital twins (DTs) in clinical trials
- •19.2 Implementation of AI in ongoing clinical trials
- •19.2.1 Integration with existing clinical trial frameworks
- •19.2.2 Quality assurance, compliance, and standardization
- •19.3 Case studies of AI in adaptive radiotherapy trials
- •19.3.1 Overview of guidance for advanced radiotherapy in clinical trials
- •19.3.2 AI in the radiotherapy clinical trial quality assurance processes
- •19.4 Ethical and regulatory considerations
- •19.4.1 Patient consent and data privacy
- •19.4.2 Bias, fairness, and transparency
- •19.4.3 Regulatory guidelines and compliance
- •19.5 Future directions and challenges
- •19.5.1 Emerging technologies and techniques
- •19.5.2 Alternative strategies
- •19.6 Conclusion
- •19.7 Summary
- •References
- •20.1 Risk management
- •20.1.1 Prospective risk assessments
- •20.1.2 Root cause analysis

Artificial Intelligence in Adaptive Radiation Therapy
[46] Qamar S et al 2019 Amide proton transfer MRI detects early changes in nasopharyngeal
carcinoma: providing a potential imaging marker for treatment response Eur. Arch.
Oto-Rhino-Laryngol.
[47] Kim C H, Lee J H, Lee J W, Kim E and Choi S-H 2022 Introducing a new biomarker named
R2*-BOLD-MRI parameter to assess treatment response in osteosarcoma J. Magn. Reson.
Imaging JMRI
[48] Jiang L, Weatherall P T, McColl R W, Tripathy D and Mason R P 2013 Blood oxygenation
level-dependent (BOLD) contrast magnetic resonance imaging (MRI) for prediction
of breast cancer chemotherapy response: a pilot study J. Magn. Reson. Imaging JMRI
1083–92
[49] Muruganandham M et al 2014 3-dimensional magnetic resonance spectroscopic imaging at 3
tesla for early response assessment of glioblastoma patients during external beam radiation
therapy Int. J. Radiat. Oncol. Biol. Phys.
[50] Nelson S J et al 2016 Serial analysis of 3D H-1 MRSI for patients with newly diagnosed
GBM treated with combination therapy that includes bevacizumab J. Neurooncol.
[51] Bezabeh T et al 2005 Prediction of treatment response in head and neck cancer by magnetic
resonance spectroscopy AJNR Am. J. Neuroradiol. 26 2108–13
[52] Bolan P J et al 2017 MR spectroscopy of breast cancer for assessing early treatment response:
results from the ACRIN 6657 MRS trial J. Magn. Reson. Imaging
[53] Kerkmeijer L G W et al 2021 Focal boost to the intraprostatic tumor in external beam
radiotherapy for patients with localized prostate cancer: results from the FLAME randomized phase III trial J. Clin. Oncol.
[54] Keall P et al 2022 ICRU REPORT 97: MRI-guided radiation therapy using MRI-linear
accelerators J. ICRU
[55] Bryant J M et al 2023 Stereotactic magnetic resonance-guided adaptive and non-adaptive
radiotherapy on combination MR-linear accelerators: current practice and future directions
Cancers
[56] Krishnan S et al 2016 Focal radiation therapy dose escalation improves overall survival in
locally advanced pancreatic cancer patients receiving induction chemotherapy and consolidative chemoradiation Int. J. Radiat. Oncol. Biol. Phys.
[57] Reyngold M et al 2021 Association of ablative radiation therapy with survival among
patients with inoperable pancreatic cancer JAMA Oncol.
[58] Bahig H et al 2018 Magnetic resonance-based response assessment and dose adaptation in
human papilloma virus positive tumors of the oropharynx treated with radiotherapy (MRADAPTOR): an R-IDEAL stage 2a-2b/Bayesian phase II trial Clin. Transl. Radiat. Oncol.
13 19–23
[59] Johnson P M et al 2022 Deep learning reconstruction enables highly accelerated bipara-
metric MR imaging of the prostate J. Magn. Reson. Imaging JMRI
[60] Afat S et al 2023 Acquisition time reduction of diffusion-weighted liver imaging using deep
learning image reconstruction Diagn. Interv. Imaging
[61] Kawamura M et al 2020; Accelerated acquisition of high-resolution diffusion-weighted
imaging of the brain with a multi-shot echo-planar sequence: deep-learning-based denoising
Magn. Reson. Med. Sci.
[62] Kaye E A et al 2020 Accelerating prostate diffusion-weighted MRI using a guided denoising
convolutional neural network: retrospective feasibility study Radiol. Artif. Intell.
15 2081
276 505–12
56 538–46
37
90 181–9
130 171–9
46 290–302
39 787–96
22 1–100
94 755–65
7 735–8
56 184–95
104 178–84
20 99–105
2 e200007
17-20

Artificial Intelligence in Adaptive Radiation Therapy
[63] Hong Y, Chen G, Yap P-T and Shen D 2019 Multifold acceleration of diffusion MRI via
deep learning reconstruction from slice-undersampled data Inf. Process. Med. Imaging Proc.
Conf.
11492 530–41
[64] Ueda T et al 2022 Deep learning reconstruction of diffusion-weighted MRI improves image
quality for prostatic imaging Radiology
[65] Albay E, Demir U and Unal G 2018 Diffusion MRI spatial super-resolution using generative
adversarial networks Predictive Intelligence in Medicine
Park (Cham: Springer International) pp 155–63
[66] Chatterjee S et al 2021 ShuffleUNet: super resolution of diffusion-weighted MRIs using deep
learning arXiv:
[67] Hu Z et al 2020 Distortion correction of single-shot EPI enabled by deep-learning
NeuroImage
[68] Ye X et al 2023 Simultaneous superresolution reconstruction and distortion correction for
single-shot EPI DWI using deep learning Magn. Reson. Med.
[69] Ulas C et al 2019 Convolutional neural networks for direct inference of pharmacokinetic
parameters: application to stroke dynamic contrast-enhanced MRI Front. Neurol.
[70] Bliesener Y, Acharya J and Nayak K S 2020 Efficient DCE-MRI parameter and uncertainty
estimation using a neural network IEEE Trans. Med. Imaging
[71] Zou J, Balter J M and Cao Y 2020 Estimation of pharmacokinetic parameters from DCE-
MRI by extracting long and short time-dependent features using an LSTM network Med.
Phys.
47 3447–57
[72] Ottens T et al 2022 Deep learning DCE-MRI parameter estimation: application in
pancreatic cancer Med. Image. Anal.
[73] Bertleff M et al 2017 Diffusion parameter mapping with the combined intravoxel incoherent
motion and kurtosis model using artificial neural networks at 3 T NMR Biomed.
[74] Barbieri S, Gurney-Champion O J, Klaassen R and Thoeny H C 2020 Deep learning how to
fit an intravoxel incoherent motion model to diffusion-weighted MRI Magn. Reson. Med.
312–21
[75] Vasylechko S D, Warfield S K, Afacan O and Kurugol S 2022 Self-supervised IVIM DWI
parameter estimation with a physics based forward model Magn. Reson. Med.
[76] Kaandorp M P T et al 2021 Improved unsupervised physics-informed deep learning for
intravoxel incoherent motion modeling and evaluation in pancreatic cancer patients Magn.
Reson. Med.
[77] Lee W, Kim B and Park H 2021 Quantification of intravoxel incoherent motion with
optimized b-values using deep neural network Magn. Reson. Med.
[78] Trebeschi S et al 2017 Deep learning for fully-automated localization and segmentation of
rectal cancer on multiparametric MR Sci. Rep.
[79] Gurney-Champion O J, Kieselmann J P, Wong K H, Ng-Cheng-Hin B, Harrington K and
Oelfke U 2020 A convolutional neural network for contouring metastatic lymph nodes on
diffusion-weighted magnetic resonance images for assessment of radiotherapy response Phys.
Imaging Radiat. Oncol.
[80] Chen Y, Xing L, Yu L, Bagshaw H P, Buyyounouski M K and Han B 2020 Automatic
intraprostatic lesion segmentation in multiparametric magnetic resonance images with
proposed multiple branch UNet Med. Phys.
2102.12898
221 117170
86 2250–65
15 1–7
303 373–81
ed I Rekik, G Unal, E Adeli and S H
89 2456–70
9 1147
39 1712–23
80 102512
30 e3833
83
87 904–14
86 230–44
7 5301
47 6421–9
17-21

Artificial Intelligence in Adaptive Radiation Therapy
[81] Liang Y et al 2020 Auto-segmentation of pancreatic tumor in multi-parametric MRI using
deep convolutional neural networks Radiother. Oncol. J. Eur. Soc. Ther. Radiol. Oncol.
193–200
[82] Nalepa J et al 2020 Fully-automated deep learning-powered system for DCE-MRI analysis
of brain tumors Artif. Intell. Med.
[83] Mazaheri Y et al 2022 Evaluation of cancer outcome assessment using MRI: a review of
deep-learning methods BJR∣Open
[84] Gao Y, Pham J, Yoon S, Cao M, Hu P and Yang Y 2021 Recent advances in functional
MRI to predict treatment response for locally advanced rectal cancer Curr. Colorectal
Cancer Rep.
[85] Gurney-Champion O J, Landry G, Redalen K R and Thorwarth D 2022 Potential of deep
learning in quantitative magnetic resonance imaging for personalized radiotherapy Semin.
Radiat. Oncol.
[86] Fu J et al 2020 Deep learning-based radiomic features for improving neoadjuvant chemo-
radiation response prediction in locally advanced rectal cancer Phys. Med. Biol.
[87] Gao Y et al 2021 A preliminary study of deep learning-based treatment response prediction
for soft tissue sarcoma using longitudinal diffusion MRI Med. Phys.
[88] Yoon J et al 2024 Added value of dynamic contrast-enhanced MR imaging in deep learning-
based prediction of local recurrence in grade 4 adult-type diffuse gliomas patients Sci. Rep.
14 2171
[89] Shukla-Dave A et al 2019 Quantitative imaging biomarkers alliance (QIBA) recommenda-
tions for improved precision of DWI and DCE-MRI derived biomarkers in multicenter
oncology trials J. Magn. Reson. Imaging JMRI
17 77–87
32 377–88
102 101769
4 20210072
48 3262–372
49 e101–21
145
65 075001
17-22

IOP Publishing
Artificial Intelligence in Adaptive Radiation Therapy
Yi Wang and X. Sharon Qi
Chapter 18
Artificial intelligence in proton adaptive
radiation therapy
Brian Winey
Adaptive proton therapy workflows are being investigated by multiple institutions
and research groups. While both photon and proton adaptive therapy workflows
must overcome numerous challenges, the unique aspects of proton interactions
resulting in energy deposition in tissue and the limited availability of image guidance
with accurate tissue composition information make adaptive proton therapy workflow development a more challenging translational endeavor. Much work has
contributed to the clinical deployment of adaptive proton therapy workflows,
mostly using in-room CT imaging which achieves the most accurate tissue decomposition. To address the outstanding challenges of adaptive proton therapy, AI tools
are being developed to address most of the components of an adaptive workflow,
including image correction, contour propagation, treatment planning, dose calculations, and QA of both imaging and treatment plans.
18.1 Proton ART
18.1.1 Clinical context and necessity
Proton therapy uses the Bragg peak physical depth dose to increase the therapeutic
ratio of dose to target versus dose to normal tissue. While the clinical significance
remains a topic of clinical trials and biological studies, the physical depth dose is well
defined in water, dependent on the initial proton kinetic energy, and multiple Bragg
peaks can be combined to deliver a prescribed dose to a volume of tissue, either from
a single beam angle or multiple beams with individualized beamlet weights.
Given the well-defined relationship between the initial kinetic energy and range of
the proton in water [1], there remain multiple sources of uncertainty of the proton
range when modeling and delivering protons of a specific energy to a patient or
phantom [2]. Some of the range uncertainties are systematic and addressed with
more precise CT stopping power ratio (SPR) image calibration, accelerator
doi:10.1088/978-0-7503-6119-4ch18 18-1 ª IOP Publishing Ltd 2025. All rights,
including for text and data mining (TDM), artificial intelligence (AI) training, and similar technologies, are reserved.

Artificial Intelligence in Adaptive Radiation Therapy
Figure 18.1. A diagram of adaptive proton therapy workflows. AI tools are being developed for Image
registration, contour propagation, dose calculation, and online image corrections. (Adapted with permission
3]. Copyright 2021 Institute of Physics and Engineering in Medicine.)
from [
commissioning, Monte Carlo dose calculations, and treatment planning margins.
For other non-systematic range uncertainties, there are proposed methods to reduce
the impacts on the dose delivery, namely margins and in vivo imaging, but adaptive
proton therapy workflows can identify changes in the patient that impact the proton
range [3]. Set-up uncertainties and anatomic changes during the course of treatment,
either daily or slower anatomic changes, can be detected in daily volumetric
imaging. There have been multiple studies of in vivo and 4D imaging to detect
higher frequency changes in the patient anatomy due to breathing motion and other
intrafractional changes [4]. Figure 18.1 provides an illustration of the adaptive
workflows being investigated for proton therapy.
Adaptive proton therapy workflows have three primary aims:
1. Detect and measure anatomic changes and set-up changes in the patient
geometry.
2. Calculate and quantify any dose differences in the target and surrounding
organs, particularly range differences.
3. Generate a new treatment plan or determine the original plan satisfies all
clinical goals.
18-2

Artificial Intelligence in Adaptive Radiation Therapy
Much work continues in all the above aims, even more so when considering real time
imaging for 4D or other intrafractional anatomic changes. Before discussing the
roles of artificial intelligence (AI) in adaptive proton therapy, it is important to
summarize briefly the patient populations along with the respective changes in the
patient anatomy and the current state of adaptive proton therapy workflows.
18.1.2 Patient populations
While there are some patient populations that are more likely to be treated with
proton therapy than other external beam or internal radiation therapy, the primary
focus of this section will be the different patient populations classified by the
geometric changes and the associated need for imaging and adaptation.
18.1.2.1 Set-up uncertainties
The daily set-up uncertainties of patient populations can vary from sub-millimeter in cranial treatments [5, 6]tolargermagnitudesfortargetsinmore
challenging locations, especially more deformable soft tissue targets. For patient
populations that have reproducible set-up with limited (< 1–2 mm) uncertainties,
including anatomic changes, there is generally not a need for adaptive workflows.
The initial planning CT is most likely a representative image of the daily patient
geometry and small uncertainties can be incorporated into the initial planning
target, either with a planning target volume (PTV) or beam specificPTV.Most
proton treatment facilities incorporate set-up and range uncertainties into the
initial treatment plan and reproducible set-ups can be taken into account in the
initial treatment plan.
Set-up uncertainties can become larger and vary with daily positioning for different
reasons. Some soft tissue targets such as sarcomas can be set up with high
reproducibility using bony anatomy or implanted fiducials as surrogates for the target
geometry including shape and location in the patient anatomy. Improved imaging
techniques such as in-room CT and the latest CBCT technologies can provide
visualization of soft tissue targets but the set-up uncertainties are generally larger
for soft tissue targets, particularly those in the thoracic and abdominal regions.
An additional reason for increased set up uncertainties is the size of the target.
Proton therapy is often used for craniospinal irradiation (CSI) and heck and neck
(H&N) primary lesions with larger nodal volumes which involve large treatment
fields covering anatomic regions that can move relative to each other. The set-up
uncertainties can be reduced with multiple isocenters and repeated imaging but the
motion of one part of the target relative to another component can increase the
set-up uncertainties.
In both previous patient populations, soft tissue and large targets, there is a
potential need for adaptive proton therapy to improve the target doses and reduce
the risk of extra dose in the neighboring OARs. The following section will begin to
unpack the implications of anatomic changes, including the impact on set-up
uncertainties.
18-3

Artificial Intelligence in Adaptive Radiation Therapy
18.1.2.2 Daily or slower anatomic changes
For some patient populations, there can be slow changes in the patient anatomy, for
example weight loss, that can reduce the efficacy of the immobilization device and
give rise to larger variations in the daily patient position. While rigid 3D and 6D
shifts can be applied to the patient position to minimize the impacts of the patient
geometry changes, the set-up uncertainties can be increased, for example in the
H&N patient population where the cranial immobilization might remain reproducible, but the neck nodal region can have increased set-up uncertainties where the
mask is less tight after patient weight loss. Not all patients within a specific primary
treatment site are subject to the same slow anatomic changes [7] but the implications
of the anatomic changes can be challenging to incorporate into the initial plan, even
when using robust optimization [8, 9], thus giving rise to the need for daily or weekly
adaptive proton therapy.
18.1.2.3 Real time anatomic changes
Real time anatomic changes are a challenge for all external beam radiation therapy
but especially impactful for proton therapy. Most common are the real time changes
due to respiratory motion. Compounding the respiratory motion with the target
moving outside the treatment volume, proton therapy also encounters changes in
tissue density as well as interplay for dynamic deliveries such as scanning deliveries.
While photon external beam treatments also encounter the same real time anatomic
changes for thoracic and abdominal targets, the impacts on proton dose distributions
are more sensitive to the real time changes. At this time, there are limited options for
proton delivery systems to adaptively address the impacts of real time anatomic
changes. Interventions that have been found to be most effective include rescanning,
gating, and breath-hold, in increasing patient intervention. For systems that include an
in-room CT, there is the ability to also perform a 4D CT or other respiratory motion
analysis in the treatment position and use this information for an adaptive workflow.
CBCT reconstruction methods are being developed to address the respiratory motion
and AI can be a helpful tool to address the CBCT motion artifacts.
Aside from respiratory motion, other real time anatomic changes can include
swallowing, eye movements, bowel changes, and bladder filling. Eye movements
during proton therapy are typically gated with direct imaging of the eye but other
real time anatomic changes during proton therapy are not regularly detected or
measured. To fully extend adaptive workflows to real time anatomic changes will
require developments of more imaging options to both detect and measure the real
time patient anatomy.
18.1.3 Imaging and adaptive workflows
18.1.3.1 Offline workflows
Imaging for radiation therapy can be divided into two categories, offline and online.
Much work has studied the role of offline imaging for adaptive proton therapy.
Some patient studies and clinical workflows recommend or require offline CT, MRI,
or PET imaging during the course of treatment to evaluate the clinical impact of the
18-4

Artificial Intelligence in Adaptive Radiation Therapy
radiation treatment as well as anatomic changes in the patient. Depending upon the
disease site, changes can include tumor growth or shrinkage, weight loss or gain, and
fluid buildup or drainage. When using offline imaging, the time required to process
the three steps of the adaptive workflow is less critical.
18.1.3.2 Online workflows
Online adaptive workflows for proton therapy are becoming more common as more
online imaging technologies are deployed in clinical proton therapy facilities. Online
imaging can either be immediately before the treatment is delivered or during the
treatment delivery. Historically, most online imaging was 2D planar imaging until
in-room or nearby CTs were deployed in some facilities such as the Paul Scherrer
Institute (PSI). Additionally, real time imaging of the PET signal was developed at
Gesellschaft für Schwerionenforschung, Helmholtz Centre for Heavy Ion Research,
and Heidelberg Ion Beam Therapy Center [4, 10–14]. These earliest measurements
of the patient anatomy and beam delivery were not used for complete online
adaptive workflows due to the time required for contour propagation, plan creation,
and dose calculations. Online adaptive workflows require rapid software applications to process the three adaptive steps.
Currently, CBCT and in-room CT imaging are available in a majority of proton
therapy facilities [15, 16], thus allowing for adaptive proton therapy workflows
based upon the available 3D imaging of the patient in the treatment position and at
isocenter for many CBCT systems. Along with the increased availability of in-room
volumetric imaging, the adaptive steps of contour propagation through rigid and
deformable registration, plan optimization, and dose calculations have each seen
tremendous improvements in speed. Still, there remains a need for improvements in
each of the steps of an adaptive proton therapy workflow which gives rise to the
opportunity for AI in adaptive proton therapy.
18.1.4 Rationale for AI in proton ART
Other chapters in this book will explore some of the common applications of AI in
radiation therapy, including applications that directly impact adaptive proton therapy.
Some of the most necessary AI developments for adaptive proton therapy address the
workflow steps that require additional time using analytic or brute force methods.
Many of the tools for adaptive proton therapy workflows are mature and ready for
clinical use but often require minutes to hours for processing. AI can dramatically
decrease the time needed for these processes. AI tools for adaptive proton therapy
workflows can be classified broadly into imaging, registration, and dose calculations.
Compared to applications of AI in adaptive photon therapy, the dosimetric
properties of the Bragg peak can impose greater constraints on the accuracy and
precision of the AI tools. For example, the image pixel accuracies for AI generated
synthetic CTs will have a greater impact on the proton dose calculations than the
photon dose calculations. Additionally, proton therapy can be used for anatomic
arrangements where a sharp gradient will spare a critical organ. The accuracy of the
relative stopping powers, the registration, the contouring, and the dose calculation
18-5

Artificial Intelligence in Adaptive Radiation Therapy
can have more pronounced impacts on the proton therapy plan optimization and
delivery. The use of AI in adaptive proton therapy workflows can greatly improve
the proton therapy delivery and it requires additional quality assurance checks to
generate confidence in the AI tools.
18.2 AI in proton ART
18.2.1 Imaging
The most development of AI for adaptive proton therapy has been focused on the
improvement of the volumetric imaging for online dose calculations. While the
current diagnostic CT quality of in-room CT imaging is generally accepted as
sufficient for proton dose calculations, the image quality of CBCT is insufficient for
dose calculations without substantial improvements. Historically, the image quality
of CBCT was addressed with simple scatter models and hardware modifications,
namely anti-scatter grids. These software and hardware modifications improved the
image quality such that registrations could be performed more accurately but the
image quality was still insufficient for dose calculations.
Analytic models were proposed to address the scatter contamination in the CBCT
projections. Such projections could more accurately predict scatter components and
mitigate the impact of scatter contamination on the reconstructed CBCT Hounsfield
units [17]. While the analytic models could improve the image quality, even to a level
sufficient for proton dose calculations [18], the time required for such analytic
model-based corrections was prohibitive for online adaptive proton therapy workflows. Additionally, the analytic models typically functioned in the projection space,
requiring access to the CBCT projection data, data that are not easily accessible in
all imaging systems.
There have been other correction methods proposed to improve the CBCT image
quality, including look up table (LUT), deformed CT, and histogram matching. The
LUT and histogram matching can be performed rapidly but fail to address all scatter
artifacts, particularly when the images have large amounts of cupping and streak
artifacts. Deformation of the CT can generate highly accurate corrected CBCT
image intensities when there are few artifacts or anatomic differences between the
floating and reference images. Deforming the CT can require a large amount of time
and fails in the presence of large artifacts and anatomic differences, particularly air
pockets [19]. To address the time required for deformation of volumetric imaging,
AI can be employed as discussed in chapter 9.
The current uses of AI can be broadly separated into image domain and
projection domain models. The advantages of the image domain corrections include
more readily accessible data and the ability to correct both the artifacts from the
scatter and the reconstruction algorithm, typically an FDK backprojection. The
projection domain corrections can more directly the patient and image specific
scatter components in the projection space. Specifically, for each unique combination of patient geometry and image system (source and panel) position, the scatter
component in the projection space will change based on the underlying physical
18-6

Artificial Intelligence in Adaptive Radiation Therapy
Figure 18.2. The use of AI to correct the online CBCT imaging is demonstrated in this image. (Adapted with
permission from [
21]. Copyright 2020 Institute of Physics and Engineering in Medicine.)
conditions. Such image and patient specific variability is not as easily determined
after reconstruction.
When considering AI applications for CBCT corrections in proton therapy,
Hansen et al [20] first published a CNN named SCATTERNET which demonstrated significant and rapid image quality improvements. Subsequent studies have
further developed, tested, and validated CNNs for the improvement of CBCT image
quality [21], as seen in fi gure 18.2. More recent studies have iterated with different
imaging systems, treatment sites, and projection domain. A CNN has been
demonstrated to be a rapid and reliable AI tool for CBCT image correction [22].
In addition to CNN models, other groups have used other AI models to correct
the CBCT image quality with GANS and cycleGANs, using both paired and
unpaired image sets. More recently, additional models such as transformers have
been translated into adaptive therapy [23]. Outside of adaptive proton therapy
research studies, there are numerous groups developing AI tools for CBCT image
quality improvements in a more diagnostic context. The proton RT groups have
many more potential models to test and validate for proton dose calculations.
18.2.2 Deformable and rigid registration
As stated above, deformable and rigid registrations are essential for adaptive proton
therapy workflows when generating deformed CTs as a surrogate for the insufficient
18-7
Соседние файлы в папке Библиотека им академика М.И. Перельмана
