Machine Learning in Radiation Oncology: Opportunities, Requirements, and Needs

Sensitivity of automated and manual treatment planning approaches to contouring variation in early-breast cancer treatment

PURPOSE

One of the advantages of integrating automated processes in treatment planning is the reduction of manual planning variability. This study aims to assess whether a deep-learning-based auto-planning solution can also reduce the contouring variation-related impact on the planned dose for early-breast cancer treatment.

METHODS

Auto- and manual plans were optimized for 20 patients using both auto- and manual OARs, including both lungs, right breast, heart, and left-anterior-descending (LAD) artery. Differences in terms of recalculated dose (ΔDrcM,ΔDrcA) and reoptimized dose (ΔDroM,ΔDroA) for manual (M) and auto (A)-plans, were evaluated on manual structures. The correlation between several geometric similarities and dose differences was also explored (Spearman's test).

RESULTS

Auto-contours were found slightly smaller in size than manual contours for right breast and heart and more than twice larger for LAD. Recalculated dose differences were found negligible for both planning approaches except for heart (ΔDrcM=-0.4 Gy, ΔDrcA=-0.3 Gy) and right breast (ΔDrcM=-1.2 Gy, ΔDrcA=-1.3 Gy) maximum dose. Re-optimized dose differences were considered equivalent to recalculated ones for both lungs and LAD, while they were significantly smaller for heart (ΔDroM=-0.2 Gy, ΔDroA=-0.2 Gy) and right breast (ΔDroM =-0.3 Gy, ΔDroA=-0.9 Gy) maximum dose. Twenty-one correlations were found for ΔDrcM,A (M=8,A=13) that reduced to four for ΔDroM,A (M=3,A=1).

CONCLUSIONS

The sensitivity of auto-planning to contouring variation was found not relevant when compared to manual planning, regardless of the method used to calculate the dose differences. Nonetheless, the method employed to define the dose differences strongly affected the correlation analysis resulting highly reduced when dose was reoptimized, regardless of the planning approach.

Advancing equitable and personalized cancer care: Novel applications and priorities of artificial intelligence for fairness and inclusivity in the patient care workflow

Patient care workflows are highly multimodal and intertwined: the intersection of data outputs provided from different disciplines and in different formats remains one of the main challenges of modern oncology. Artificial Intelligence (AI) has the potential to revolutionize the current clinical practice of oncology owing to advancements in digitalization, database expansion, computational technologies, and algorithmic innovations that facilitate discernment of complex relationships in multimodal data. Within oncology, radiation therapy (RT) represents an increasingly complex working procedure, involving many labor-intensive and operator-dependent tasks. In this context, AI has gained momentum as a powerful tool to standardize treatment performance and reduce inter-observer variability in a time-efficient manner. This review explores the hurdles associated with the development, implementation, and maintenance of AI platforms and highlights current measures in place to address them. In examining AI's role in oncology workflows, we underscore that a thorough and critical consideration of these challenges is the only way to ensure equitable and unbiased care delivery, ultimately serving patients' survival and quality of life.

A new architecture combining convolutional and transformer-based networks for automatic 3D multi-organ segmentation on CT images

PURPOSE

Deep learning-based networks have become increasingly popular in the field of medical image segmentation. The purpose of this research was to develop and optimize a new architecture for automatic segmentation of the prostate gland and normal organs in the pelvic, thoracic, and upper gastro-intestinal (GI) regions.

METHODS

We developed an architecture which combines a shifted-window (Swin) transformer with a convolutional U-Net. The network includes a parallel encoder, a cross-fusion block, and a CNN-based decoder to extract local and global information and merge related features on the same scale. A skip connection is applied between the cross-fusion block and decoder to integrate low-level semantic features. Attention gates (AGs) are integrated within the CNN to suppress features in image background regions. Our network is termed "SwinAttUNet." We optimized the architecture for automatic image segmentation. Training datasets consisted of planning-CT datasets from 300 prostate cancer patients from an institutional database and 100 CT datasets from a publicly available dataset (CT-ORG). Images were linearly interpolated and resampled to a spatial resolution of (1.0 × 1.0× 1.5) mm3 . A volume patch (192 × 192 × 96) was used for training and inference, and the dataset was split into training (75%), validation (10%), and test (15%) cohorts. Data augmentation transforms were applied consisting of random flip, rotation, and intensity scaling. The loss function comprised Dice and cross-entropy equally weighted and summed. We evaluated Dice coefficients (DSC), 95th percentile Hausdorff Distances (HD95), and Average Surface Distances (ASD) between results of our network and ground truth data.

RESULTS

SwinAttUNet, DSC values were 86.54 ± 1.21, 94.15 ± 1.17, and 87.15 ± 1.68% and HD95 values were 5.06 ± 1.42, 3.16 ± 0.93, and 5.54 ± 1.63 mm for the prostate, bladder, and rectum, respectively. Respective ASD values were 1.45 ± 0.57, 0.82 ± 0.12, and 1.42 ± 0.38 mm. For the lung, liver, kidneys and pelvic bones, respective DSC values were: 97.90 ± 0.80, 96.16 ± 0.76, 93.74 ± 2.25, and 89.31 ± 3.87%. Respective HD95 values were: 5.13 ± 4.11, 2.73 ± 1.19, 2.29 ± 1.47, and 5.31 ± 1.25 mm. Respective ASD values were: 1.88 ± 1.45, 1.78 ± 1.21, 0.71 ± 0.43, and 1.21 ± 1.11 mm. Our network outperformed several existing deep learning approaches using only attention-based convolutional or Transformer-based feature strategies, as detailed in the results section.

CONCLUSIONS

We have demonstrated that our new architecture combining Transformer- and convolution-based features is able to better learn the local and global context for automatic segmentation of multi-organ, CT-based anatomy.

Predictive modeling of dose-volume parameters of carcinoma tongue cases using machine learning models

The aim of this study is to create a single institution-based machine learning model for a dose prediction generation tool for post-operative carcinoma of the tongue cases prospectively. Intensity-modulated radiotherapy (IMRT) plans for 20 patients with carcinoma of the tongue were generated using the Eclipse treatment planning system. A machine learning model was generated using a Python 3.10 computer language in a Jupyter notebook using Anaconda software. The PTVs and OARs doses obtained from the clinical treatment plans were used as a primary dataset. Machine learning models are built with two different datasets (10 and 20) for each selected volume. Volumes from 10 new sets of patients were fed into the software for predicting the corresponding dose values. Through the input given, the plan generated dose values of 10 patients were compared with the predicted outcomes of the 10 and 20 dataset models. The model created using the PTVs volume data predicted the dose values with increased accuracy. By verifying the model prediction with the TPS generated value, both the 10 and 20 dataset models predict all the 10 PTVs data within an error bound of 3% and most of the OARs data within an error bound of 5%. The dosimetric features implemented in the machine learning models reasonably predict both the PTVs dose parameter and OARs constraints and give confidence in decision-making during the clinical planning process.

The influence of artificial intelligence on the work of the medical physicist in radiotherapy practice: a short review

There have been many applications and influences of Artificial intelligence (AI) in many sectors and its professionals, that of radiotherapy and the medical physicist is no different. AI and technological advances have necessitated changing roles of medical physicists due to the development of modernized technology with image-guided accessories for the radiotherapy treatment of cancer patients. Given the changing role of medical physicists in ensuring patient safety and optimal care, AI can reshape radiotherapy practice now and in some years to come. Medical physicists' roles in radiotherapy practice have evolved to meet technology for the management of better patient care in the age of modern radiotherapy. This short review provides an insight into the influence of AI on the changing role of medical physicists in each specific chain of the workflow in radiotherapy in which they are involved.

Clinical implementation of deep learning-based automated left breast simultaneous integrated boost radiotherapy treatment planning

Background and purpose

Automation in radiotherapy treatment planning aims to improve both the quality and the efficiency of the process. The aim of this study was to report on a clinical implementation of a Deep Learning (DL) auto-planning model for left-sided breast cancer.

Materials and methods

The DL model was developed for left-sided breast simultaneous integrated boost treatments under deep-inspiration breath-hold. Eighty manual dose distributions were revised and used for training. Ten patients were used for model validation. The model was then used to design 17 clinical auto-plans. Manual and auto-plans were scored on a list of clinical goals for both targets and organs-at-risk (OARs). For validation, predicted and mimicked dose (PD and MD, respectively) percent error (PE) was calculated with respect to manual dose. Clinical and validation cohorts were compared in terms of MD only.

Results

Median values of both PD and MD validation plans fulfilled the evaluation criteria. PE was < 1% for targets for both PD and MD. PD was well aligned to manual dose while MD left lung mean dose was significantly less (median:5.1 Gy vs 6.1 Gy). The left-anterior-descending artery maximum dose was found out of requirements (median values:+5.9 Gy and + 2.9 Gy, for PD and MD respectively) in three validation cases, while it was reduced for clinical cases (median:-1.9 Gy). No other clinically significant differences were observed between clinical and validation cohorts.

Conclusion

Small OAR differences observed during the model validation were not found clinically relevant. The clinical implementation outcomes confirmed the robustness of the model.

Translation of AI into oncology clinical practice

Artificial intelligence (AI) is a transformative technology that is capturing popular imagination and can revolutionize biomedicine. AI and machine learning (ML) algorithms have the potential to break through existing barriers in oncology research and practice such as automating workflow processes, personalizing care, and reducing healthcare disparities. Emerging applications of AI/ML in the literature include screening and early detection of cancer, disease diagnosis, response prediction, prognosis, and accelerated drug discovery. Despite this excitement, only few AI/ML models have been properly validated and fewer have become regulated products for routine clinical use. In this review, we highlight the main challenges impeding AI/ML clinical translation. We present different clinical use cases from the domains of radiology, radiation oncology, immunotherapy, and drug discovery in oncology. We dissect the unique challenges and opportunities associated with each of these cases. Finally, we summarize the general requirements for successful AI/ML implementation in the clinic, highlighting specific examples and points of emphasis including the importance of multidisciplinary collaboration of stakeholders, role of domain experts in AI augmentation, transparency of AI/ML models, and the establishment of a comprehensive quality assurance program to mitigate risks of training bias and data drifts, all culminating toward safer and more beneficial AI/ML applications in oncology labs and clinics.

Africa's readiness for artificial intelligence in clinical radiotherapy delivery: Medical physicists to lead the way

BACKGROUND

There have been several proposals by researchers for the introduction of Artificial Intelligence (AI) technology due to its promising role in radiotherapy practice. However, prior to the introduction of the technology, there are certain general recommendations that must be achieved. Also, the current challenges of AI must be addressed. In this review, we assess how Africa is prepared for the integration of AI technology into radiotherapy service delivery.

METHODS

To assess the readiness of Africa for integration of AI in radiotherapy services delivery, a narrative review of the available literature from PubMed, Science Direct, Google Scholar, and Scopus was conducted in the English language using search terms such as Artificial Intelligence, Radiotherapy in Africa, Machine Learning, Deep Learning, and Quality Assurance.

RESULTS

We identified a number of issues that could limit the successful integration of AI technology into radiotherapy practice. The major issues include insufficient data for training and validation of AI models, lack of educational curriculum for AI radiotherapy-related courses, no/limited AI teaching professionals, funding, and lack of AI technology and resources. Solutions identified to facilitate smooth implementation of the technology into radiotherapy practices within the region include: creating an accessible national data bank, integrating AI radiotherapy training programs into Africa's educational curriculum, investing in AI technology and resources such as electronic health records and cloud storage, and creation of legal laws and policies to support the use of the technology. These identified solutions need to be implemented on the background of creating awareness among health workers within the radiotherapy space.

CONCLUSION

The challenges identified in this review are common among all the geographical regions in the African continent. Therefore, all institutions offering radiotherapy education and training programs, management of the medical centers for radiotherapy and oncology, national and regional professional bodies for medical physics, ministries of health, governments, and relevant stakeholders must take keen interest and work together to achieve this goal.

Evaluation treatment planning system for oropharyngeal cancer patient using machine learning

Oropharyngeal cancer (OPC) comprises a group of various malignant tumours that grow in the throat, larynx, mouth, sinuses, and nose. THE RESEARCH AIMS: to investigate the performance of the OPC VMAT model by comparison to clinical plans in terms of dosimetric parameters and normal tissue complication probabilities.

PURPOSE

Tune the model which at least matches the performance of clinical created photon treatment plans and analyse and find the most appropriate strategic plan scheme for OPC.

METHODS AND MATERIALS

The machine learning (ML) plans are compared to the reference plans (clinical plans) based on dose constraints and target coverage. VMAT oropharynx ML model of Raystation development 11B version (non-clinical) was used. A model was trained by using different modalities. A different strategy of machine learning and clinical plans was performed for five patients. The dose Prescribed for OPC is 70 Gy, 2 Gy per fraction (2Gy/Fx). The PTV was derived for the primary tumour and secondary tumour, PTV+7000 cGy and PTV_5425 cGy volumetric modulated arc therapy (VMAT) were used with beams performing a full 360° rotation around the single isocenter.

RESULTS

Organs at risk were observed that the volume of L-Eye in clinical plan (AF) for the case1 treatment planning could be successfully used ensuring efficiency and lower than MLVMAT and MLVMAT-org plans were 372 cGy, 697 cGy and 667 cGy respectively, while showed case2, case3, case4 and case5 are better to protect the critical organs in ML plan compare with a clinical plan. DHI for the PTV-7000 and PTV-5425 is between 1 and 1.34, While DCI for PTV-7000 and PTV-5425 is between 0.98 and 1.

Development and multi-institutional validation of a convolutional neural network to detect vertebral body mis-alignments in 2D x-ray setup images

BACKGROUND

Misalignment to the incorrect vertebral body remains a rare but serious patient safety risk in image-guided radiotherapy (IGRT).

PURPOSE

Our group has proposed that an automated image-review algorithm be inserted into the IGRT process as an interlock to detect off-by-one vertebral body errors. This study presents the development and multi-institutional validation of a convolutional neural network (CNN)-based approach for such an algorithm using patient image data from a planar stereoscopic x-ray IGRT system.

METHODS

X-rays and digitally reconstructed radiographs (DRRs) were collected from 429 spine radiotherapy patients (1592 treatment fractions) treated at six institutions using a stereoscopic x-ray image guidance system. Clinically-applied, physician approved, alignments were used for true-negative, "no-error" cases. "Off-by-one vertebral body" errors were simulated by translating DRRs along the spinal column using a semi-automated method. A leave-one-institution-out approach was used to estimate model accuracy on data from unseen institutions as follows: All of the images from five of the institutions were used to train a CNN model from scratch using a fixed network architecture and hyper-parameters. The size of this training set ranged from 5700 to 9372 images, depending on exactly which five institutions were contributing data. The training set was randomized and split using a 75/25 split into the final training/ validation sets. X-ray/ DRR image pairs and the associated binary labels of "no-error" or "shift" were used as the model input. Model accuracy was evaluated using images from the sixth institution, which were left out of the training phase entirely. This test set ranged from 180 to 3852 images, again depending on which institution had been left out of the training phase. The trained model was used to classify the images from the test set as either "no-error" or "shifted", and the model predictions were compared to the ground truth labels to assess the model accuracy. This process was repeated until each institution's images had been used as the testing dataset.

RESULTS

When the six models were used to classify unseen image pairs from the institution left out during training, the resulting receiver operating characteristic area under the curve values ranged from 0.976 to 0.998. With the specificity fixed at 99%, the corresponding sensitivities ranged from 61.9% to 99.2% (mean: 77.6%). With the specificity fixed at 95%, sensitivities ranged from 85.5% to 99.8% (mean: 92.9%).

CONCLUSION

This study demonstrated the CNN-based vertebral body misalignment model is robust when applied to previously unseen test data from an outside institution, indicating that this proposed additional safeguard against misalignment is feasible.

Correlation between patient-specific quality assurance in volumetric modulated arc therapy and 2D dose image features

In radiotherapy, air-filled ion chamber detectors are ubiquitously used in routine dose measurements for treatment planning. However, its use has been restricted by intrinsic low spatial resolution barriers. We developed one procedure for patient-specific quality assurance (QA) in arc radiotherapy by coalescing two adjacent measurement images into a single image to improve spatial resolution and sampling frequency, and investigated how different spatial resolutions affect the QA results. PTW 729 and 1500 ion chamber detectors were used for dosimetric verification via coalescing two measurements with 5 mm-couch shift and the isocenter, and only isocenter measurement, which we call coalescence and standard acquisition (SA). Statistical process control (SPC), process capability analysis (PCA), and receiver operating characteristic (ROC) curve were used to compare the performance of the two procedures in determining tolerance levels and identifying clinically relevant errors. By analyzing 1256 γ values calculated on interpolated data points, our results indicated that detector 1500 showed higher averages in coalescence cohorts at different tolerance criteria and the dispersion degrees were spread out smaller. Detector 729 yielded a slightly lower process capability of 0.79, 0.76, 1.10, and 1.34, but detector 1500 exhibited somewhat different results of 0.94, 1.42, 1.19, and 1.60 in magnitude. The results of SPC individual control chart showed that cases in coalescence cohorts with γ values lowering its lower control limit (LCL) were greater than those in SA cohorts for detector 1500. A combination of the width of multi-leaf collimator (MLC) leaf, the cross-sectional area of the single detector, and the spacing between adjacent detectors might lead to discrepancies in percent γ values across diverse spatial resolution scenarios. The accuracy of reconstructed volume dose is mainly determined by the interpolation algorithm used in dosimetric systems. The magnitude of filling factor in the ion chamber detectors determined its ability to detect dose deviations. SPC and PCA results indicated that coalescence procedure could detect more potential failure QA results than SA while enhancing action thresholds.

Personalizing digital pain management with adapted machine learning approach

Introduction

Digital therapeutics (DT) emerged and has been expanding rapidly for pain management. However, the efficacy of such approaches demonstrates substantial heterogeneity. Machine learning (ML) approaches provide a great opportunity for personalizing the efficacy of DT. However, the ML model accuracy is mainly associated with reduced clinical interpretability. Moreover, classical ML models are not adapted for the longitudinal nature of the DT follow-up data, which may also include nonlinear fluctuations.

Objectives

This study presents an analytical framework for personalized pain management using piecewise mixed-effects model trees, considering the data dependencies, nonlinear trajectories, and boosting model interpretability.

Methods

We demonstrated the implementation of the model with posture biofeedback training data of 3610 users collected during 8 weeks. The users reported their pain levels and posture quality. We developed personalized models for nonlinear time-related fluctuations of pain levels, posture quality, and weekly training duration using age, gender, and body mass index as potential moderating factors.

Results

Pain levels and posture quality demonstrated strong improvement during the first 3 weeks of the training, followed by a sustained pattern. The age of the users moderated the time fluctuations in pain levels, whereas age and gender interactively moderated the trajectories in the posture quality. Train duration increased during the first 3 weeks only for older users, whereas all the users decreased the training duration during the next 5 weeks.

Conclusions

This analytical framework offers an opportunity for investigating the personalized efficacy of digital therapeutics for pain management, taking into account users' characteristics and boosting interpretability and can benefit from including more users' characteristics.

Spot delivery error predictions for intensity modulated proton therapy using robustness analysis with machine learning

The purpose of this work is to assess the robustness of treatment plans when spot delivery errors were predicted with a machine learning (ML) model for intensity modulated proton therapy (IMPT). Over 6000 machine log files from delivered IMPT treatment plans were included in this study. From these log files, over 4.1  × $ \times \ $ 10⁶ delivered proton spots were used to train the ML model. The presented model was tested and used to predict the spot position as well as the monitor units (MU) per spot, based on the original planning parameters. Two patient plans (one accelerated partial breast irradiation [APBI] and one ependymoma) were recalculated with the predicted spot position/MUs by the ML model and then were re-analyzed for robustness. Plans with ML predicted spots were less robust than the original clinical plans. In the APBI plan, dosimetric changes to the left lung and heart were not clinically relevant. In the ependymoma plan, the hot spot in the brainstem decreased and the hot spot in the cervical cord increased. Despite these differences, after robustness analysis, both ML spot delivery error plans resulted in >95% of the CTV receiving >95% of the prescription dose. The presented workflow has the potential benefit of including realistic spots information for plan quality checks in IMPT. This work demonstrates that in the two example plans, the plans were still robust when accounting for spot delivery errors as predicted by the ML model.

Deep learning-based prediction of deliverable adaptive plans for MR-guided adaptive radiotherapy: A feasibility study

Purpose

Fast and automated plan generation is desirable in radiation therapy (RT), in particular, for MR-guided online adaptive RT (MRgOART) or real-time (intrafractional) adaptive RT (MRgRART), to reduce replanning time. The purpose of this study is to investigate the feasibility of using deep learning to quickly predict deliverable adaptive plans based on a target dose distribution for MRgOART/MRgRART.

Methods

A conditional generative adversarial network (cGAN) was trained to predict the MLC leaf sequence corresponding to a target dose distribution based on reference plan created prior to MRgOART using a 1.5T MR-Linac. The training dataset included 50 ground truth dose distributions and corresponding beam parameters (aperture shapes and weights) created during MRgOART for 10 pancreatic cancer patients (each with five fractions). The model input was the dose distribution from each individual beam and the output was the predicted corresponding field segments with specific shape and weight. Patient-based leave-one-out-cross-validation was employed and for each model trained, four (44 training beams) out of five fractionated plans of the left-out patient were set aside for testing purposes. We deliberately kept a single fractionated plan in the training dataset so that the model could learn to replan the patient based on a prior plan. The model performance was evaluated by calculating the gamma passing rate of the ground truth dose vs. the dose from the predicted adaptive plan and calculating max and mean dose metrics.

Results

The average gamma passing rate (95%, 3mm/3%) among 10 test cases was 88%. In general, we observed 95% of the prescription dose to PTV achieved with an average 7.6% increase of max and mean dose, respectively, to OARs for predicted replans. Complete adaptive plans were predicted in ≤20 s using a GTX 1660TI GPU.

Conclusion

We have proposed and demonstrated a deep learning method to generate adaptive plans automatically and rapidly for MRgOART. With further developments using large datasets and the inclusion of patient contours, the method may be implemented to accelerate MRgOART process or even to facilitate MRgRART.

Prediction of portal dosimetry quality assurance results using log files-derived errors and machine learning techniques

Objective

This work aims to use machine learning models to predict gamma passing rate of portal dosimetry quality assurance with log file derived features. This allows daily treatment monitoring for patients and reduce wear and tear on EPID detectors to save cost and prevent downtime.

Methods

578 VMAT trajectory log files selected from prostate, lung and spine SBRT were used in this work. Four machine learning models were explored to identify the best performing regression model for predicting gamma passing rate within each sub-site and the entire unstratified data. Predictors used in these models comprised of hand-crafted log file-derived features as well as modulation complexity score. Cross validation was used to evaluate the model performance in terms of R² and RMSE.

Result

Using gamma passing rate of 1%/1mm criteria and entire dataset, LASSO regression has a R² of 0.121 ± 0.005 and RMSE of 4.794 ± 0.013%, SVM regression has a R² of 0.605 ± 0.036 and RMSE of 3.210 ± 0.145%, Random Forest regression has a R² of 0.940 ± 0.019 and RMSE of 1.233 ± 0.197%. XGBoost regression has the best performance with a R² and RMSE value of 0.981 ± 0.015 and 0.652 ± 0.276%, respectively.

Conclusion

Log file-derived features can predict gamma passing rate of portal dosimetry with an average error of less than 2% using the 1%/1mm criteria. This model can potentially be applied to predict the patient specific QA results for every treatment fraction.

Automatic quality assurance of radiotherapy treatment plans using Bayesian networks: A multi-institutional study

Purpose

Artificial intelligence applications in radiation oncology have been the focus of study in the last decade. The introduction of automated and intelligent solutions for routine clinical tasks, such as treatment planning and quality assurance, has the potential to increase safety and efficiency of radiotherapy. In this work, we present a multi-institutional study across three different institutions internationally on a Bayesian network (BN)-based initial plan review assistive tool that alerts radiotherapy professionals for potential erroneous or suboptimal treatment plans.

Methods

Clinical data were collected from the oncology information systems in three institutes in Europe (Maastro clinic - 8753 patients treated between 2012 and 2020) and the United States of America (University of Vermont Medical Center [UVMMC] - 2733 patients, University of Washington [UW] - 6180 patients, treated between 2018 and 2021). We trained the BN model to detect potential errors in radiotherapy treatment plans using different combinations of institutional data and performed single-site and cross-site validation with simulated plans with embedded errors. The simulated errors consisted of three different categories: i) patient setup, ii) treatment planning and iii) prescription. We also compared the strategy of using only diagnostic parameters or all variables as evidence for the BN. We evaluated the model performance utilizing the area under the receiver-operating characteristic curve (AUC).

Results

The best network performance was observed when the BN model is trained and validated using the dataset in the same center. In particular, the testing and validation using UVMMC data has achieved an AUC of 0.92 with all parameters used as evidence. In cross-validation studies, we observed that the BN model performed better when it was trained and validated in institutes with similar technology and treatment protocols (for instance, when testing on UVMMC data, the model trained on UW data achieved an AUC of 0.84, compared with an AUC of 0.64 for the model trained on Maastro data). Also, combining training data from larger clinics (UW and Maastro clinic) and using it on smaller clinics (UVMMC) leads to satisfactory performance with an AUC of 0.85. Lastly, we found that in general the BN model performed better when all variables are considered as evidence.

Conclusion

We have developed and validated a Bayesian network model to assist initial treatment plan review using multi-institutional data with different technology and clinical practices. The model has shown good performance even when trained on data from clinics with divergent profiles, suggesting that the model is able to adapt to different data distributions.

Machine learning-based predictions of gamma passing rates for virtual specific-plan verification based on modulation maps, monitor unit profiles, and composite dose images

Machine learning (ML) methods have been implemented in radiotherapy to aid virtual specific-plan verification protocols, predicting gamma passing rates (GPR) based on calculated modulation complexity metrics because of their direct relation to dose deliverability. Nevertheless, these metrics might not comprehensively represent the modulation complexity, and automatically extracted features from alternative predictors associated with modulation complexity are needed. For this reason, three convolutional neural networks (CNN) based models were trained to predict GPR values (regression and classification), using respectively three predictors: (1) the modulation maps (MM) from the multi-leaf collimator, (2) the relative monitor units per control point profile (MUcp), and (3) the composite dose image (CDI) used for portal dosimetry, from 1024 anonymized prostate plans. The models' performance was assessed for classification and regression by the area under the receiver operator characteristic curve (AUC_ROC) and Spearman's correlation coefficient (r). Finally, four hybrid models were designed using all possible combinations of the three predictors. The prediction performance for the CNN-models using single predictors (MM, MUcp, and CDI) were AUC_ROC = 0.84 ± 0.03, 0.77 ± 0.07, 0.75 ± 0.04, andr= 0.6, 0.5, 0.7. Contrastingly, the hybrid models (MM + MUcp, MM + CDI, MUcp+CDI, MM + MUcp+CDI) performance were AUC_ROC = 0.94 ± 0.03, 0.85 ± 0.06, 0.89 ± 0.06, 0.91 ± 0.03, andr= 0.7, 0.5, 0.6, 0.7. The MP, MUcp, and CDI are suitable predictors for dose deliverability models implementing ML methods. Additionally, hybrid models are susceptible to improving their prediction performance, including two or more input predictors.

A deep image-to-image network organ segmentation algorithm for radiation treatment planning: principles and evaluation

Background

We describe and evaluate a deep network algorithm which automatically contours organs at risk in the thorax and pelvis on computed tomography (CT) images for radiation treatment planning.

Methods

The algorithm identifies the region of interest (ROI) automatically by detecting anatomical landmarks around the specific organs using a deep reinforcement learning technique. The segmentation is restricted to this ROI and performed by a deep image-to-image network (DI2IN) based on a convolutional encoder-decoder architecture combined with multi-level feature concatenation. The algorithm is commercially available in the medical products “syngo.via RT Image Suite VB50” and “AI-Rad Companion Organs RT VA20” (Siemens Healthineers). For evaluation, thoracic CT images of 237 patients and pelvic CT images of 102 patients were manually contoured following the Radiation Therapy Oncology Group (RTOG) guidelines and compared to the DI2IN results using metrics for volume, overlap and distance, e.g., Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD₉₅). The contours were also compared visually slice by slice.

Results

We observed high correlations between automatic and manual contours. The best results were obtained for the lungs (DSC 0.97, HD₉₅ 2.7 mm/2.9 mm for left/right lung), followed by heart (DSC 0.92, HD₉₅ 4.4 mm), bladder (DSC 0.88, HD₉₅ 6.7 mm) and rectum (DSC 0.79, HD₉₅ 10.8 mm). Visual inspection showed excellent agreements with some exceptions for heart and rectum.

Conclusions

The DI2IN algorithm automatically generated contours for organs at risk close to those by a human expert, making the contouring step in radiation treatment planning simpler and faster. Few cases still required manual corrections, mainly for heart and rectum.

A digital physician peer to automatically detect erroneous prescriptions in radiotherapy

Appropriate dosing of radiation is crucial to patient safety in radiotherapy. Current quality assurance depends heavily on a physician peer-review process, which includes a review of the treatment plan's dose and fractionation. Potentially, physicians may not identify errors during this manual peer review due to time constraints and caseload. A novel prescription anomaly detection algorithm is designed that utilizes historical data from the past to predict anomalous cases. Such a tool can serve as an electronic peer who will assist the peer-review process providing extra safety to the patients. In our primary model, we create two dissimilarity metrics, R and F. R defining how far a new patient's prescription is from historical prescriptions. F represents how far away a patient's feature set is from that of the group with an identical or similar prescription. We flag prescription if either metric is greater than specific optimized cut-off values. We use thoracic cancer patients (n = 2504) as an example and extracted seven features. Our testing set f1 score is between 73%-94% for different treatment technique groups. We also independently validate our results by conducting a mock peer review with three thoracic specialists. Our model has a lower type II error rate compared to the manual peer-review by physicians.

Towards a safe and efficient clinical implementation of machine learning in radiation oncology by exploring model interpretability, explainability and data-model dependency

The interest in machine learning (ML) has grown tremendously in recent years, partly due to the performance leap that occurred with new techniques of deep learning, convolutional neural networks for images, increased computational power, and wider availability of large datasets. Most fields of medicine follow that popular trend and, notably, radiation oncology is one of those that are at the forefront, with already a long tradition in using digital images and fully computerized workflows. ML models are driven by data, and in contrast with many statistical or physical models, they can be very large and complex, with countless generic parameters. This inevitably raises two questions, namely, the tight dependence between the models and the datasets that feed them, and the interpretability of the models, which scales with its complexity. Any problems in the data used to train the model will be later reflected in their performance. This, together with the low interpretability of ML models, makes their implementation into the clinical workflow particularly difficult. Building tools for risk assessment and quality assurance of ML models must involve then two main points: interpretability and data-model dependency. After a joint introduction of both radiation oncology and ML, this paper reviews the main risks and current solutions when applying the latter to workflows in the former. Risks associated with data and models, as well as their interaction, are detailed. Next, the core concepts of interpretability, explainability, and data-model dependency are formally defined and illustrated with examples. Afterwards, a broad discussion goes through key applications of ML in workflows of radiation oncology as well as vendors' perspectives for the clinical implementation of ML.

The feasibility study on the generalization of deep learning dose prediction model for volumetric modulated arc therapy of cervical cancer

Purpose

To develop a 3D‐Unet dose prediction model to predict the three‐dimensional dose distribution of volumetric modulated arc therapy (VMAT) for cervical cancer and test the dose prediction performance of the model in endometrial cancer to explore the feasibility of model generalization.

Methods

One hundred and seventeen cases of cervical cancer and 20 cases of endometrial cancer treated with VMAT were used for the model training, validation, and test. The prescribed dose was 50.4 Gy in 28 fractions. Eight independent channels of contoured structures were input to the model, and the dose distribution was used as the output of the model. The 3D‐Unet prediction model was trained and validated on the training set (n = 86) and validation set (n = 11), respectively. Then the model was tested on the test set (n = 20) of cervical cancer and endometrial cancer, respectively. The results between clinical dose distribution and predicted dose distribution were compared in the following aspects: (a) the mean absolute error (MAE) within the body, (b) the Dice similarity coefficients (DSCs) under different isodose volumes, (c) the dosimetric indexes including the mean dose (D_mean), the received dose of 2 cm³ (D_2cc), the percentage volume of receiving 40 Gy dose of organs‐at‐risk (V₄₀), planning target volume (PTV) D_98%, and homogeneity index (HI), (d) dose–volume histograms (DVHs).

Results

The model can accurately predict the dose distribution of the VMAT plan for cervical cancer and endometrial cancer. The overall average MAE and maximum MAE for cervical cancer were 2.43 ± 3.17% and 3.16 ± 4.01% of the prescribed dose, respectively, and for endometrial cancer were 2.70 ± 3.54% and 3.85 ± 3.11%. The average DSCs under different isodose volumes is above 0.9. The predicted dosimetric indexes and DVHs are equivalent to the clinical dose for both cervical cancer and endometrial cancer, and there is no statistically significant difference.

Conclusion

A 3D‐Unet dose prediction model was developed for VMAT of cervical cancer, which can predict the dose distribution accurately for cervical cancer. The model can also be generalized for endometrial cancer with good performance.

Regression Analysis of Rectal Cancer and Possible Application of Artificial Intelligence (AI) Utilization in Radiotherapy

Artificial Intelligence (AI) has been widely employed in the medical field in recent years in such areas as image segmentation, medical image registration, and computer-aided detection. This study explores one application of using AI in adaptive radiation therapy treatment planning by predicting the tumor volume reduction rate (TVRR). Cone beam computed tomography (CBCT) scans of twenty rectal cancer patients were collected to observe the change in tumor volume over the course of a standard five-week radiotherapy treatment. In addition to treatment volume, patient data including patient age, gender, weight, number of treatment fractions, and dose per fraction were also collected. Application of a stepwise regression model showed that age, dose per fraction and weight were the best predictors for tumor volume reduction rate.

Machine Learning for Head and Neck Cancer: A Safe Bet?-A Clinically Oriented Systematic Review for the Radiation Oncologist

BACKGROUND AND PURPOSE

Machine learning (ML) is emerging as a feasible approach to optimize patients' care path in Radiation Oncology. Applications include autosegmentation, treatment planning optimization, and prediction of oncological and toxicity outcomes. The purpose of this clinically oriented systematic review is to illustrate the potential and limitations of the most commonly used ML models in solving everyday clinical issues in head and neck cancer (HNC) radiotherapy (RT).

MATERIALS AND METHODS

Electronic databases were screened up to May 2021. Studies dealing with ML and radiomics were considered eligible. The quality of the included studies was rated by an adapted version of the qualitative checklist originally developed by Luo et al. All statistical analyses were performed using R version 3.6.1.

RESULTS

Forty-eight studies (21 on autosegmentation, four on treatment planning, 12 on oncological outcome prediction, 10 on toxicity prediction, and one on determinants of postoperative RT) were included in the analysis. The most common imaging modality was computed tomography (CT) (40%) followed by magnetic resonance (MR) (10%). Quantitative image features were considered in nine studies (19%). No significant differences were identified in global and methodological scores when works were stratified per their task (i.e., autosegmentation).

DISCUSSION AND CONCLUSION

The range of possible applications of ML in the field of HN Radiation Oncology is wide, albeit this area of research is relatively young. Overall, if not safe yet, ML is most probably a bet worth making.

Reduction of inter-observer differences in the delineation of the target in spinal metastases SBRT using an automatic contouring dedicated system

BACKGROUND

Approximately one third of cancer patients will develop spinal metastases, that can be associated with back pain, neurological symptoms and deterioration in performance status. Stereotactic radiosurgery (SRS) and stereotactic body radiotherapy (SBRT) have been offered in clinical practice mainly for the management of oligometastatic and oligoprogressive patients, allowing the prescription of high total dose delivered in one or few sessions to small target volumes, minimizing the dose exposure of normal tissues. Due to the high delivered doses and the proximity of critical organs at risk (OAR) such as the spinal cord, the correct definition of the treatment volume becomes even more important in SBRT treatment, thus making it necessary to standardize the method of target definition and contouring, through the adoption of specific guidelines and specific automatic contouring tools. An automatic target contouring system for spine SBRT is useful to reduce inter-observer differences in target definition. In this study, an automatic contouring tool was evaluated.

METHODS

Simulation CT scans and MRI data of 20 patients with spinal metastases were evaluated. To evaluate the advantage of the automatic target contouring tool (Elements SmartBrush Spine), which uses the identification of different densities within the target vertebra, we evaluated the agreement of the contours of 20 spinal target (2 cervical, 9 dorsal and 9 lumbar column), outlined by three independent observers using the automatic tool compared to the contours obtained manually, and measured by DICE similarity coefficient.

RESULTS

The agreement of GTV contours outlined by independent operators was superior with the use of the automatic contour tool compared to manually outlined contours (mean DICE coefficient 0.75 vs 0.57, p = 0.048).

CONCLUSIONS

The dedicated contouring tool allows greater precision and reduction of inter-observer differences in the delineation of the target in SBRT spines. Thus, the evaluated system could be useful in the setting of spinal SBRT to reduce uncertainties of contouring increasing the level of precision on target delivered doses.

Reconstruction of volume averaging effect-free continuous photon beam profiles from discrete ionization chamber array measurements using a machine learning technique

PURPOSE

The use of the ionization chamber array ICProfiler (ICP) is limited by its relatively poor detector spatial resolution and the inherent volume averaging effect (VAE). The purpose of this work is to study the feasibility of reconstructing VAE-free continuous photon beam profiles from ICP measurements with a machine learning technique.

METHODS

In- and cross-plane photon beam profiles of a 6 MV beam from an Elekta linear accelerator, ranging from 2 × 2 to 10 × 10 cm² at 1.5 cm, 5 cm, and 10 cm depth, were measured with an ICP. The discrete measurements were interpolated with a Makima method to obtain continuous beam profiles. Artificial neural networks (ANNs) were trained to restore the penumbra of the beam profiles. Plane-specific (in- and cr-plane) ANNs and a combined ANN were separately trained. The performance of the ANNs was evaluated using the penumbra width difference (PWD, the difference between the penumbra widths of the reconstructed and the reference profile). The plane-specific and the combined ANNs were compared to study the feasibility of using a single ANN for both in- and cross-plane.

RESULTS

The profiles reconstructed with all the ANNs had excellent agreement with the reference. For in-plane, the ANNs reduced the PWD from 1.6 ± 0.7 mm at 1.5 cm depth to 0.1 ± 0.1 mm, from 1.8 ± 0.6 mm at 5.0 cm depth to 0.1 ± 0.1 mm, and from 2.4 ± 0.1 mm at 10.0 cm depth to 0.0 ± 0.0 mm; for cross-plane, the ANNs reduced the PWD from 1.2 ± 0.4 mm at 1.5 cm depth, 1.2 ± 0.3 mm at 5.0 cm depth, and 1.6 ± 0.1 mm at 10.0 cm depth, to 0.1 ± 0.1 mm.

CONCLUSIONS

This study demonstrated the feasibility of using simple ANNs to reconstruct VAE-free continuous photon beam profiles from discrete ICP measurements. A combined ANN can restore the penumbra of in- and cross-plane beam profiles of various fields at different depths.

Deep learning methods to generate synthetic CT from MRI in radiotherapy: A literature review

PURPOSE

In radiotherapy, MRI is used for target volume and organs-at-risk delineation for its superior soft-tissue contrast as compared to CT imaging. However, MRI does not provide the electron density of tissue necessary for dose calculation. Several methods of synthetic-CT (sCT) generation from MRI data have been developed for radiotherapy dose calculation. This work reviewed deep learning (DL) sCT generation methods and their associated image and dose evaluation, in the context of MRI-based dose calculation.

METHODS

We searched the PubMed and ScienceDirect electronic databases from January 2010 to March 2021. For each paper, several items were screened and compiled in figures and tables.

RESULTS

This review included 57 studies. The DL methods were either generator-only based (45% of the reviewed studies), or generative adversarial network (GAN) architecture and its variants (55% of the reviewed studies). The brain and pelvis were the most commonly investigated anatomical localizations (39% and 28% of the reviewed studies, respectively), and more rarely, the head-and-neck (H&N) (15%), abdomen (10%), liver (5%) or breast (3%). All the studies performed an image evaluation of sCTs with a diversity of metrics, with only 36 studies performing dosimetric evaluations of sCT.

CONCLUSIONS

The median mean absolute errors were around 76 HU for the brain and H&N sCTs and 40 HU for the pelvis sCTs. For the brain, the mean dose difference between the sCT and the reference CT was <2%. For the H&N and pelvis, the mean dose difference was below 1% in most of the studies. Recent GAN architectures have advantages compared to generator-only, but no superiority was found in term of image or dose sCT uncertainties. Key challenges of DL-based sCT generation methods from MRI in radiotherapy is the management of movement for abdominal and thoracic localizations, the standardization of sCT evaluation, and the investigation of multicenter impacts.

Artificial intelligence in brachytherapy: a summary of recent developments

Artificial intelligence (AI) applications, in the form of machine learning and deep learning, are being incorporated into practice in various aspects of medicine, including radiation oncology. Ample evidence from recent publications explores its utility and future use in external beam radiotherapy. However, the discussion on its role in brachytherapy is sparse. This article summarizes available current literature and discusses potential uses of AI in brachytherapy, including future directions. AI has been applied for brachytherapy procedures during almost all steps, starting from decision-making till treatment completion. AI use has led to improvement in efficiency and accuracy by reducing the human errors and saving time in certain aspects. Apart from direct use in brachytherapy, AI also contributes to contemporary advancements in radiology and associated sciences that can affect brachytherapy decisions and treatment. There is a renewal of interest in brachytherapy as a technique in recent years, contributed largely by the understanding that contemporary advances such as intensity modulated radiotherapy and stereotactic external beam radiotherapy cannot match the geometric gains and conformality of brachytherapy, and the integrated efforts of international brachytherapy societies to promote brachytherapy training and awareness. Use of AI technologies may consolidate it further by reducing human effort and time. Prospective validation over larger studies and incorporation of AI technologies for a larger patient population would help improve the efficiency and acceptance of brachytherapy. The enthusiasm favoring AI needs to be balanced against the short duration and quantum of experience with AI in limited patient subsets, need for constant learning and re-learning to train the AI algorithms, and the inevitability of humans having to take responsibility for the correctness and safety of treatments.

Artificial intelligence in radiation oncology

Artificial intelligence (AI) is a computer science that tries to mimic human-like intelligence in machines that use computer software and algorithms to perform specific tasks without direct human input. Machine learning (ML) is a subunit of AI that uses data-driven algorithms that learn to imitate human behavior based on a previous example or experience. Deep learning is an ML technique that uses deep neural networks to create a model. The growth and sharing of data, increasing computing power, and developments in AI have initiated a transformation in healthcare. Advances in radiation oncology have produced a significant amount of data that must be integrated with computed tomography imaging, dosimetry, and imaging performed before each fraction. Of the many algorithms used in radiation oncology, has advantages and limitations with different computational power requirements. The aim of this review is to summarize the radiotherapy (RT) process in workflow order by identifying specific areas in which quality and efficiency can be improved by ML. The RT stage is divided into seven stages: patient evaluation, simulation, contouring, planning, quality control, treatment application, and patient follow-up. A systematic evaluation of the applicability, limitations, and advantages of AI algorithms has been done for each stage.

Convolutional neural network and transfer learning for dose volume histogram prediction for prostate cancer radiotherapy

To adopt a transfer learning approach and establish a convolutional neural network (CNN) model for the prediction of rectum and bladder dose-volume histograms (DVH) in prostate patients treated with a VMAT technique. One hundred forty-four VMAT patients with intermediate or high-risk prostate cancer were included in this study. Data were split into two sets: 120 and 24 patients, respectively. The second set was used for final validation. To ensure the accuracy of the training data, we developed a ground-truth analysis for detecting and correcting for all potential outliers. We used transfer learning in combination with a pre-trained VGG-16 network. We dropped the fully connected layers from the VGG-16 and added a new fully connected neural network. The inputs for the CNN were a 2D image of the volumes contoured in the CT, but we only retained the geometrical information of every CT-slice. The outputs were the corresponding rectum and bladder DVH for every slice. We used a confusion matrix to analyze the performance of our model. Our model achieved 100% and 81% of true positive and true negative predictions, respectively. We have an overall accuracy of 87.5%, a misclassification rate of 12.5%, and a precision of 100%. We have successfully developed a model for reliable prediction of rectum and bladder DVH in prostate patients by applying a previously pre-trained CNN. To our knowledge, this is the first attempt to apply transfer learning to the prediction of DVHs that accounts for the ground truth problem.

Ethical evaluation of artificial intelligence applications in radiotherapy using the Four Topics Approach

Artificial Intelligence is the capability of a machine to imitate intelligent human behavior. An important impact can be expected from Artificial Intelligence throughout the workflow of radiotherapy (such as automated organ segmentation, treatment planning, prediction of outcome and quality assurance). However, ethical concerns regarding the binding agreement between the patient and the physician have followed the introduction of artificial intelligence. Through the recording of personal and social moral values in addition to the usual demographics and the implementation of these as distinctive inputs to matching algorithms, ethical concerns such as consistency, applicability and relevance can be solved. In the meantime, physicians' awareness of the ethical dimension in their decision-making should be challenged, so that they prioritize treating their patients and not diseases, remain vigilant to preserve patient safety, avoid unintended harm and establish institutional policies on these issues.

National Cancer Institute Workshop on Artificial Intelligence in Radiation Oncology: Training the Next Generation

PURPOSE

Artificial intelligence (AI) is about to touch every aspect of radiation therapy, from consultation to treatment planning, quality assurance, therapy delivery, and outcomes modeling. There is an urgent need to train radiation oncologists and medical physicists in data science to help shepherd AI solutions into clinical practice. Poorly trained personnel may do more harm than good when attempting to apply rapidly developing and complex technologies. As the amount of AI research expands in our field, the radiation oncology community needs to discuss how to educate future generations in this area.

METHODS AND MATERIALS

The National Cancer Institute (NCI) Workshop on AI in Radiation Oncology (Shady Grove, MD, April 4-5, 2019) was the first of 2 data science workshops in radiation oncology hosted by the NCI in 2019. During this workshop, the Training and Education Working Group was formed by volunteers among the invited attendees. Its members represent radiation oncology, medical physics, radiology, computer science, industry, and the NCI.

RESULTS

In this perspective article written by members of the Training and Education Working Group, we provide and discuss action points relevant for future trainees interested in radiation oncology AI: (1) creating AI awareness and responsible conduct; (2) implementing a practical didactic curriculum; (3) creating a publicly available database of training resources; and (4) accelerating learning and funding opportunities.

CONCLUSION

Together, these action points can facilitate the translation of AI into clinical practice.

Practical considerations for prostate hypofractionation in the developing world

External beam radiotherapy is an effective curative treatment option for localized prostate cancer, the most common cancer in men worldwide. However, conventionally fractionated courses of curative external beam radiotherapy are usually 8-9 weeks long, resulting in a substantial burden to patients and the health-care system. This problem is exacerbated in low-income and middle-income countries where health-care resources might be scarce and patient funds limited. Trials have shown a clinical equipoise between hypofractionated schedules of radiotherapy and conventionally fractionated treatments, with the advantage of drastically shortening treatment durations with the use of hypofractionation. The hypofractionated schedules are supported by modern consensus guidelines for implementation in clinical practice. Furthermore, several economic evaluations have shown improved cost effectiveness of hypofractionated therapy compared with conventional schedules. However, these techniques demand complex infrastructure and advanced personnel training. Thus, a number of practical considerations must be borne in mind when implementing hypofractionation in low-income and middle-income countries, but the potential gain in the treatment of this patient population is substantial.

Auto-segmentations by convolutional neural network in cervical and anorectal cancer with clinical structure sets as the ground truth

•

The network provided auto-segmentations with high overlap with ground truth volumes.

•

Evaluation of femoral heads/bladder auto-segmentations showed highest overlap.

•

Annotated structure sets from daily clinical practice is feasible as ground truth.

Background

It is time-consuming for oncologists to delineate volumes for radiotherapy treatment in computer tomography (CT) images. Automatic delineation based on image processing exists, but with varied accuracy and moderate time savings. Using convolutional neural network (CNN), delineations of volumes are faster and more accurate. We have used CTs with the annotated structure sets to train and evaluate a CNN.

Material and methods

The CNN is a standard segmentation network modified to minimize memory usage. We used CTs and structure sets from 75 cervical cancers and 191 anorectal cancers receiving radiation therapy at Skåne University Hospital 2014-2018. Five structures were investigated: left/right femoral heads, bladder, bowel bag, and clinical target volume of lymph nodes (CTVNs). Dice score and mean surface distance (MSD) (mm) evaluated accuracy, and one oncologist qualitatively evaluated auto-segmentations.

Results

Median Dice/MSD scores for anorectal cancer: 0.91–0.92/1.93–1.86 femoral heads, 0.94/2.07 bladder, and 0.83/6.80 bowel bag. Median Dice scores for cervical cancer were 0.93–0.94/1.42–1.49 femoral heads, 0.84/3.51 bladder, 0.88/5.80 bowel bag, and 0.82/3.89 CTVNs. With qualitative evaluation, performance on femoral heads and bladder auto-segmentations was mostly excellent, but CTVN auto-segmentations were not acceptable to a larger extent.

Discussion

It is possible to train a CNN with high overlap using structure sets as ground truth. Manually delineated pelvic volumes from structure sets do not always strictly follow volume boundaries and are sometimes inaccurately defined, which leads to similar inaccuracies in the CNN output. More data that is consistently annotated is needed to achieve higher CNN accuracy and to enable future clinical implementation.

Morphological autoencoders for apnea detection in respiratory gating radiotherapy

BACKGROUND AND OBJECTIVE

Respiratory gating training is a common technique to increase patient proprioception, with the goal of (e.g.) minimizing the effects of organ motion during radiotherapy. In this work, we devise a system based on autoencoders for classification of regular, apnea and unconstrained breathing patterns (i.e. multiclass).

METHODS

Our approach is based on morphological analysis of the respiratory signals, using an autoencoder trained on regular breathing. The correlation between the input and output of the autoencoder is used to train and test several classifiers in order to select the best. Our approach is evaluated in a novel real-world respiratory gating biofeedback training dataset and on the Apnea-ECG reference dataset.

RESULTS

Accuracies of 95 ± 3.5% and 87 ± 6.6% were obtained for two different datasets, in the classification of breathing and apnea. These results suggest the viability of a generalised model to characterise the breathing patterns under study.

CONCLUSIONS

Using autoencoders to learn respiratory gating training patterns allows a data-driven approach to feature extraction, by focusing only on the signal's morphology. The proposed system is prone to be used in real-time and could potentially be transferred to other domains.

Machine learning applications in radiation oncology: Current use and needs to support clinical implementation

BACKGROUND AND PURPOSE

The use of artificial intelligence (AI)/ machine learning (ML) applications in radiation oncology is emerging, however no clear guidelines on commissioning of ML-based applications exist. The purpose of this study was therefore to investigate the current use and needs to support implementation of ML-based applications in routine clinical practice.

MATERIALS AND METHODS

A survey was conducted among medical physicists in radiation oncology, consisting of four parts: clinical applications (1), model training, acceptance and commissioning (2), quality assurance (QA) in clinical practice and General Data Protection Regulation (GDPR) (3), and need for education and guidelines (4). Survey answers of medical physicists of the same radiation oncology centre were treated as a separate unique responder in case reporting on different AI applications.

RESULTS

In total, 213 medical physicists from 202 radiation oncology centres were included in the analysis. Sixty-nine percent (1 4 7) was using (37%) or preparing (32%) to use ML in clinic, mostly for contouring and treatment planning. In 86%, human observers were still involved in daily clinical use for quality check of the output of the ML algorithm. Knowledge on ethics, legislation and data sharing was limited and scattered among responders. Besides the need for (implementation) guidelines, training of medical physicists and larger databases containing multicentre data was found to be the top priority to accommodate the further introduction of ML in clinical practice.

CONCLUSION

The results of this survey indicated the need for education and guidelines on the implementation and quality assurance of ML-based applications to benefit clinical introduction.

Integration of AI and Machine Learning in Radiotherapy QA

The use of machine learning and other sophisticated models to aid in prediction and decision making has become widely popular across a breadth of disciplines. Within the greater diagnostic radiology, radiation oncology, and medical physics communities promising work is being performed in tissue classification and cancer staging, outcome prediction, automated segmentation, treatment planning, and quality assurance as well as other areas. In this article, machine learning approaches are explored, highlighting specific applications in machine and patient-specific quality assurance (QA). Machine learning can analyze multiple elements of a delivery system on its performance over time including the multileaf collimator (MLC), imaging system, mechanical and dosimetric parameters. Virtual Intensity-Modulated Radiation Therapy (IMRT) QA can predict passing rates using different measurement techniques, different treatment planning systems, and different treatment delivery machines across multiple institutions. Prediction of QA passing rates and other metrics can have profound implications on the current IMRT process. Here we cover general concepts of machine learning in dosimetry and various methods used in virtual IMRT QA, as well as their clinical applications.

Overview of artificial intelligence-based applications in radiotherapy: Recommendations for implementation and quality assurance

Artificial Intelligence (AI) is currently being introduced into different domains, including medicine. Specifically in radiation oncology, machine learning models allow automation and optimization of the workflow. A lack of knowledge and interpretation of these AI models can hold back wide-spread and full deployment into clinical practice. To facilitate the integration of AI models in the radiotherapy workflow, generally applicable recommendations on implementation and quality assurance (QA) of AI models are presented. For commonly used applications in radiotherapy such as auto-segmentation, automated treatment planning and synthetic computed tomography (sCT) the basic concepts are discussed in depth. Emphasis is put on the commissioning, implementation and case-specific and routine QA of AI models needed for a methodical introduction in clinical practice.

Artificial intelligence in radiation oncology

Artificial intelligence (AI) has the potential to fundamentally alter the way medicine is practised. AI platforms excel in recognizing complex patterns in medical data and provide a quantitative, rather than purely qualitative, assessment of clinical conditions. Accordingly, AI could have particularly transformative applications in radiation oncology given the multifaceted and highly technical nature of this field of medicine with a heavy reliance on digital data processing and computer software. Indeed, AI has the potential to improve the accuracy, precision, efficiency and overall quality of radiation therapy for patients with cancer. In this Perspective, we first provide a general description of AI methods, followed by a high-level overview of the radiation therapy workflow with discussion of the implications that AI is likely to have on each step of this process. Finally, we describe the challenges associated with the clinical development and implementation of AI platforms in radiation oncology and provide our perspective on how these platforms might change the roles of radiotherapy medical professionals.

AUTOMATIC BRAIN ORGAN SEGMENTATION WITH 3D FULLY CONVOLUTIONAL NEURAL NETWORK FOR RADIATION THERAPY TREATMENT PLANNING

3D organ contouring is an essential step in radiation therapy treatment planning for organ dose estimation as well as for optimizing plans to reduce organs-at-risk doses. Manual contouring is time-consuming and its inter-clinician variability adversely affects the outcomes study. Such organs also vary dramatically on sizes - up to two orders of magnitude difference in volumes. In this paper, we present BrainSegNet, a novel 3D fully convolutional neural network (FCNN) based approach for automatic segmentation of brain organs. BrainSegNet takes a multiple resolution paths approach and uses a weighted loss function to solve the major challenge of the large variability in organ sizes. We evaluated our approach with a dataset of 46 Brain CT image volumes with corresponding expert organ contours as reference. Compared with those of LiviaNet and V-Net, BrainSegNet has a superior performance in segmenting tiny or thin organs, such as chiasm, optic nerves, and cochlea, and outperforms these methods in segmenting large organs as well. BrainSegNet can reduce the manual contouring time of a volume from an hour to less than two minutes, and holds high potential to improve the efficiency of radiation therapy workflow.

Machine learning in quantitative PET: A review of attenuation correction and low-count image reconstruction methods

The rapid expansion of machine learning is offering a new wave of opportunities for nuclear medicine. This paper reviews applications of machine learning for the study of attenuation correction (AC) and low-count image reconstruction in quantitative positron emission tomography (PET). Specifically, we present the developments of machine learning methodology, ranging from random forest and dictionary learning to the latest convolutional neural network-based architectures. For application in PET attenuation correction, two general strategies are reviewed: 1) generating synthetic CT from MR or non-AC PET for the purposes of PET AC, and 2) direct conversion from non-AC PET to AC PET. For low-count PET reconstruction, recent deep learning-based studies and the potential advantages over conventional machine learning-based methods are presented and discussed. In each application, the proposed methods, study designs and performance of published studies are listed and compared with a brief discussion. Finally, the overall contributions and remaining challenges are summarized.

Introduction to machine and deep learning for medical physicists

Recent years have witnessed tremendous growth in the application of machine learning (ML) and deep learning (DL) techniques in medical physics. Embracing the current big data era, medical physicists equipped with these state-of-the-art tools should be able to solve pressing problems in modern radiation oncology. Here, a review of the basic aspects involved in ML/DL model building, including data processing, model training, and validation for medical physics applications is presented and discussed. Machine learning can be categorized based on the underlying task into supervised learning, unsupervised learning, or reinforcement learning; each of these categories has its own input/output dataset characteristics and aims to solve different classes of problems in medical physics ranging from automation of processes to predictive analytics. It is recognized that data size requirements may vary depending on the specific medical physics application and the nature of the algorithms applied. Data processing, which is a crucial step for model stability and precision, should be performed before training the model. Deep learning as a subset of ML is able to learn multilevel representations from raw input data, eliminating the necessity for hand crafted features in classical ML. It can be thought of as an extension of the classical linear models but with multilayer (deep) structures and nonlinear activation functions. The logic of going "deeper" is related to learning complex data structures and its realization has been aided by recent advancements in parallel computing architectures and the development of more robust optimization methods for efficient training of these algorithms. Model validation is an essential part of ML/DL model building. Without it, the model being developed cannot be easily trusted to generalize to unseen data. Whenever applying ML/DL, one should keep in mind, according to Amara's law, that humans may tend to overestimate the ability of a technology in the short term and underestimate its capability in the long term. To establish ML/DL role into standard clinical workflow, models considering balance between accuracy and interpretability should be developed. Machine learning/DL algorithms have potential in numerous radiation oncology applications, including automatizing mundane procedures, improving efficiency and safety of auto-contouring, treatment planning, quality assurance, motion management, and outcome predictions. Medical physicists have been at the frontiers of technology translation into medicine and they ought to be prepared to embrace the inevitable role of ML/DL in the practice of radiation oncology and lead its clinical implementation.

Genomics models in radiotherapy: From mechanistic to machine learning

Machine learning (ML) provides a broad framework for addressing high-dimensional prediction problems in classification and regression. While ML is often applied for imaging problems in medical physics, there are many efforts to apply these principles to biological data toward questions of radiation biology. Here, we provide a review of radiogenomics modeling frameworks and efforts toward genomically guided radiotherapy. We first discuss medical oncology efforts to develop precision biomarkers. We next discuss similar efforts to create clinical assays for normal tissue or tumor radiosensitivity. We then discuss modeling frameworks for radiosensitivity and the evolution of ML to create predictive models for radiogenomics.

Prediction of the output factor using machine and deep learning approach in uniform scanning proton therapy

PURPOSE

The purpose of this work is to develop machine and deep learning-based models to predict output and MU based on measured patient quality assurance (QA) data in uniform scanning proton therapy (USPT).

METHODS

This study involves 4,231 patient QA measurements conducted over the last 6 years. In the current approach, output and MU are predicted by an empirical model (EM) based on patient treatment plan parameters. In this study, two MATLAB-based machine and deep learning algorithms - Gaussian process regression (GPR) and shallow neural network (SNN) - were developed. The four parameters from patient QA (range, modulation, field size, and measured output factor) were used to train these algorithms. The data were randomized with a training set containing 90% and a testing set containing remaining 10% of the data. The model performance during training was accessed using root mean square error (RMSE) and R-squared values. The trained model was used to predict output based on the three input parameters: range, modulation, and field size. The percent difference was calculated between the predicted and measured output factors. The number of data sets required to make prediction accuracy of GPR and SNN models' invariable was also evaluated.

RESULTS

The prediction accuracy of machine and deep learning algorithms is higher than the EM. The output predictions with [GPR, SNN, and EM] within ± 2% and ± 3% difference were [97.16%, 97.64%, and 92.95%] and [99.76%, 99.29%, and 97.18%], respectively. The GPR model outperformed the SNN with a smaller number of training data sets.

CONCLUSION

The GPR and SNN models outperformed the EM in terms of prediction accuracy. Machine and deep learning algorithms predicted the output factor and MU for USPT with higher predictive accuracy than EM. In our clinic, these models have been adopted as a secondary check of MU or output factors.