Applications and limitations of machine learning in radiation oncology

Fully automated radiotherapy treatment planning: A scan to plan challenge

BACKGROUND AND PURPOSE

Over the past decade, tools for automation of various sub-tasks in radiotherapy planning have been introduced, such as auto-contouring and auto-planning. The purpose of this study was to benchmark what degree of automation is possible.

MATERIALS AND METHODS

A challenge to perform automated treatment planning for prostate and prostate bed radiotherapy was set up. Participants were provided with simulation CTs and a treatment prescription and were asked to use automated tools to produce a deliverable radiotherapy treatment plan with as little human intervention as possible. Plans were scored for their adherence to the protocol when assessed using consensus expert contours.

RESULTS

Thirteen entries were received. The top submission adhered to 81.8% of the minimum objectives across all cases using the consensus contour, meeting all objectives in one of the ten cases. The same system met 89.5% of objectives when assessed with their own auto-contours, meeting all objectives in four of the ten cases. The majority of systems used in the challenge had regulatory clearance (Auto-contouring: 82.5%, Auto-planning: 77%). Despite the 'hard' rule that participants should not check or edit contours or plans, 69% reported looking at their results before submission.

CONCLUSIONS

Automation of the full planning workflow from simulation CT to deliverable treatment plan is possible for prostate and prostate bed radiotherapy. While many generated plans were found to require none or minor adjustment to be regarded as clinically acceptable, the result indicated there is still a lack of trust in such systems preventing full automation.

Artificial intelligence in interventional radiotherapy (brachytherapy): Enhancing patient-centered care and addressing patients' needs

This review explores the integration of artificial intelligence (AI) in interventional radiotherapy (IRT), emphasizing its potential to streamline workflows and enhance patient care. Through a systematic analysis of 78 relevant papers spanning from 2002 to 2024, we identified significant advancements in contouring, treatment planning, outcome prediction, and quality assurance. AI-driven approaches offer promise in reducing procedural times, personalizing treatments, and improving treatment outcomes for oncological patients. However, challenges such as clinical validation and quality assurance protocols persist. Nonetheless, AI presents a transformative opportunity to optimize IRT and meet evolving patient needs.

Pharmacogenomic scores in psychiatry: systematic review of current evidence

In the past two decades, significant progress has been made in the development of polygenic scores (PGSs). One specific application of PGSs is the development and potential use of pharmacogenomic- scores (PGx-scores) to identify patients who can benefit from a specific medication or are likely to experience side effects. This systematic review comprehensively evaluates published PGx-score studies in psychiatry and provides insights into their potential clinical use and avenues for future development. A systematic literature search was conducted across PubMed, EMBASE, and Web of Science databases until 22 August 2023. This review included fifty-three primary studies, of which the majority (69.8%) were conducted using samples of European ancestry. We found that over 90% of PGx-scores in psychiatry have been developed based on psychiatric and medical diagnoses or trait variants, rather than pharmacogenomic variants. Among these PGx-scores, the polygenic score for schizophrenia (PGSSCZ) has been most extensively studied in relation to its impact on treatment outcomes (32 publications). Twenty (62.5%) of these studies suggest that individuals with higher PGSSCZ have negative outcomes from psychotropic treatment - poorer treatment response, higher rates of treatment resistance, more antipsychotic-induced side effects, or more psychiatric hospitalizations, while the remaining studies did not find significant associations. Although PGx-scores alone accounted for at best 5.6% of the variance in treatment outcomes (in schizophrenia treatment resistance), together with clinical variables they explained up to 13.7% (in bipolar lithium response), suggesting that clinical translation might be achieved by including PGx-scores in multivariable models. In conclusion, our literature review found that there are still very few studies developing PGx-scores using pharmacogenomic variants. Research with larger and diverse populations is required to develop clinically relevant PGx-scores, using biology-informed and multi-phenotypic polygenic scoring approaches, as well as by integrating clinical variables with these scores to facilitate their translation to psychiatric practice.

Radiomics for clinical decision support in radiation oncology

Radiomics is a promising tool for the development of quantitative biomarkers to support clinical decision-making. It has been shown to improve the prediction of response to treatment and outcome in different settings, particularly in the field of radiation oncology by optimising the dose delivery solutions and reducing the rate of radiation-induced side effects, leading to a fully personalised approach. Despite the promising results offered by radiomics at each of these stages, standardised methodologies, reproducibility and interpretability of results are still lacking, limiting the potential clinical impact of these tools. In this review, we briefly describe the principles of radiomics and the most relevant applications of radiomics at each stage of cancer management in the framework of radiation oncology. Furthermore, the integration of radiomics into clinical decision support systems is analysed, defining the challenges and offering possible solutions for translating radiomics into a clinically applicable tool.

Quantifying impairment and disease severity using AI models trained on healthy subjects

Automatic assessment of impairment and disease severity is a key challenge in data-driven medicine. We propose a framework to address this challenge, which leverages AI models trained exclusively on healthy individuals. The COnfidence-Based chaRacterization of Anomalies (COBRA) score exploits the decrease in confidence of these models when presented with impaired or diseased patients to quantify their deviation from the healthy population. We applied the COBRA score to address a key limitation of current clinical evaluation of upper-body impairment in stroke patients. The gold-standard Fugl-Meyer Assessment (FMA) requires in-person administration by a trained assessor for 30-45 minutes, which restricts monitoring frequency and precludes physicians from adapting rehabilitation protocols to the progress of each patient. The COBRA score, computed automatically in under one minute, is shown to be strongly correlated with the FMA on an independent test cohort for two different data modalities: wearable sensors (ρ = 0.814, 95% CI [0.700,0.888]) and video (ρ = 0.736, 95% C.I [0.584, 0.838]). To demonstrate the generalizability of the approach to other conditions, the COBRA score was also applied to quantify severity of knee osteoarthritis from magnetic-resonance imaging scans, again achieving significant correlation with an independent clinical assessment (ρ = 0.644, 95% C.I [0.585,0.696]).

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.

Automation and artificial intelligence in radiation therapy treatment planning

Automation and artificial intelligence (AI) is already possible for many radiation therapy planning and treatment processes with the aim of improving workflows and increasing efficiency in radiation oncology departments. Currently, AI technology is advancing at an exponential rate, as are its applications in radiation oncology. This commentary highlights the way AI has begun to impact radiation therapy treatment planning and looks ahead to potential future developments in this space. Historically, radiation therapist's (RT's) role has evolved alongside the adoption of new technology. In Australia, RTs have key clinical roles in both planning and treatment delivery and have been integral in the implementation of automated solutions for both areas. They will need to continue to be informed, to adapt and to transform with AI technologies implemented into clinical practice in radiation oncology departments. RTs will play an important role in how AI-based automation is implemented into practice in Australia, ensuring its application can truly enable personalised and higher-quality treatment for patients. To inform and optimise utilisation of AI, research should not only focus on clinical outcomes but also AI's impact on professional roles, responsibilities and service delivery. Increased efficiencies in the radiation therapy workflow and workforce need to maintain safe improvements in practice and should not come at the cost of creativity, innovation, oversight and safety.

Evaluating capabilities of large language models: Performance of GPT-4 on surgical knowledge assessments

BACKGROUND

Artificial intelligence has the potential to dramatically alter health care by enhancing how we diagnose and treat disease. One promising artificial intelligence model is ChatGPT, a general-purpose large language model trained by OpenAI. ChatGPT has shown human-level performance on several professional and academic benchmarks. We sought to evaluate its performance on surgical knowledge questions and assess the stability of this performance on repeat queries.

METHODS

We evaluated the performance of ChatGPT-4 on questions from the Surgical Council on Resident Education question bank and a second commonly used surgical knowledge assessment, referred to as Data-B. Questions were entered in 2 formats: open-ended and multiple-choice. ChatGPT outputs were assessed for accuracy and insights by surgeon evaluators. We categorized reasons for model errors and the stability of performance on repeat queries.

RESULTS

A total of 167 Surgical Council on Resident Education and 112 Data-B questions were presented to the ChatGPT interface. ChatGPT correctly answered 71.3% and 67.9% of multiple choice and 47.9% and 66.1% of open-ended questions for Surgical Council on Resident Education and Data-B, respectively. For both open-ended and multiple-choice questions, approximately two-thirds of ChatGPT responses contained nonobvious insights. Common reasons for incorrect responses included inaccurate information in a complex question (n = 16, 36.4%), inaccurate information in a fact-based question (n = 11, 25.0%), and accurate information with circumstantial discrepancy (n = 6, 13.6%). Upon repeat query, the answer selected by ChatGPT varied for 36.4% of questions answered incorrectly on the first query; the response accuracy changed for 6/16 (37.5%) questions.

CONCLUSION

Consistent with findings in other academic and professional domains, we demonstrate near or above human-level performance of ChatGPT on surgical knowledge questions from 2 widely used question banks. ChatGPT performed better on multiple-choice than open-ended questions, prompting questions regarding its potential for clinical application. Unique to this study, we demonstrate inconsistency in ChatGPT responses on repeat queries. This finding warrants future consideration including efforts at training large language models to provide the safe and consistent responses required for clinical application. Despite near or above human-level performance on question banks and given these observations, it is unclear whether large language models such as ChatGPT are able to safely assist clinicians in providing care.

A convolutional neural network-based system for fully automatic segmentation of whole-body [68Ga]Ga-PSMA PET images in prostate cancer

PURPOSE

The aim of this study was development and evaluation of a fully automated tool for the detection and segmentation of mPCa lesions in whole-body [68Ga]Ga-PSMA-11 PET scans by using a nnU-Net framework.

METHODS

In this multicenter study, a cohort of 412 patients from three different center with all indication of PCa who underwent [68Ga]Ga-PSMA-11 PET/CT were enrolled. Two hundred cases of center 1 dataset were used for training the model. A fully 3D convolutional neural network (CNN) is proposed which is based on the self-configuring nnU-Net framework. A subset of center 1 dataset and cases of center 2 and center 3 were used for testing of model. The performance of the segmentation pipeline that was developed was evaluated by comparing the fully automatic segmentation mask with the manual segmentation of the corresponding internal and external test sets in three levels including patient-level scan classification, lesion-level detection, and voxel-level segmentation. In addition, for comparison of PET-derived quantitative biomarkers between automated and manual segmentation, whole-body PSMA tumor volume (PSMA-TV) and total lesions PSMA uptake (TL-PSMA) were calculated.

RESULTS

In terms of patient-level classification, the model achieved an accuracy of 83%, sensitivity of 92%, PPV of 77%, and NPV of 91% for the internal testing set. For lesion-level detection, the model achieved an accuracy of 87-94%, sensitivity of 88-95%, PPV of 98-100%, and F1-score of 93-97% for all testing sets. For voxel-level segmentation, the automated method achieved average values of 65-70% for DSC, 72-79% for PPV, 53-58% for IoU, and 62-73% for sensitivity in all testing sets. In the evaluation of volumetric parameters, there was a strong correlation between the manual and automated measurements of PSMA-TV and TL-PSMA for all centers.

CONCLUSIONS

The deep learning networks presented here offer promising solutions for automatically segmenting malignant lesions in prostate cancer patients using [68Ga]Ga-PSMA PET. These networks achieve a high level of accuracy in whole-body segmentation, as measured by the DSC and PPV at the voxel level. The resulting segmentations can be used for extraction of PET-derived quantitative biomarkers and utilized for treatment response assessment and radiomic studies.

A comprehensive review of dose limits, triage systems and measurement tools for consequence management of nuclear and radiological emergencies

During a radiological or nuclear emergency, occupational workers, members of the public, and emergency responders may be exposed to radionuclides, whether external or internal, through inhalation, ingestion, or wounds. In the case of internalized radiation exposure, prompt assessment of contamination is necessary to inform subsequent medical interventions. This review assembles the constituent considerations for managing nuclear and radiological incidents, focused on a parallel analysis of the evolution of radiation dose limits - notably in the emergency preparedness and response realm - alongside a discussion of triage systems and in vivo radionuclide detection tools. The review maps the development of international and national standards and regulations concerning radiation dose limits, illuminating how past incidents and accumulated knowledge have informed present emergency preparedness and response practices, specifically for internalized radiation. Additionally, the objectives and levels of radiation triage systems are explored in-depth, along with a global survey of practices and protocols. Finally, this review also focuses on in vivo detection systems and their capacities for radionuclide identification, prioritizing internalized gamma-emitting isotopes due to their broader relevance. Collectively, this study comprehensively addresses the intricacies of triage management following radiation emergencies, emphasizing the imperative for enhanced standardization and continued research in this critical domain.

Predictive model of positive surgical margins after radical prostatectomy based on Bayesian network analysis

Objective

This study aimed to analyze the independent risk factors for marginal positivity after radical prostatectomy and to evaluate the clinical value of the predictive model based on Bayesian network analysis.

Methods

We retrospectively analyzed the clinical data from 238 patients who had undergone radical prostatectomy, between June 2018 and May 2022. The general clinical data, prostate specific antigen (PSA)-derived indicators, puncture factors, and magnetic resonance imaging (MRI) characteristics were included as predictive variables, and univariate and multivariate analyses were conducted. We established a nomogram model based on the independent predictors and adopted BayesiaLab software to generate tree-augmented naive (TAN) and naive Bayesian models based on 15 predictor variables.

Results

Of the 238 patients included in the study, 103 exhibited positive surgical margins. Univariate analysis revealed that PSA density (PSAD) (P = 0.02), Gleason scores for biopsied tissue (P = 0.002) and the ratio of positive biopsy cores (P < 0.001), preoperative T staging (P < 0.001), and location of abnormal signals (P = 0.002) and the side of the abnormal signal (P = 0.009) were all statistically significant. The area under curve (AUC) of the established nomogram model based on independent predictors was 73.80%, the AUC of the naive Bayesian model based on 15 predictors was 82.71%, and the AUC of the TAN Bayesian model was 80.80%.

Conclusion

The predictive model of positive resection margin after radical prostatectomy based on Bayesian network demonstrated high accuracy and usefulness.

Technical Note: Generalizable and Promptable Artificial Intelligence Model to Augment Clinical Delineation in Radiation Oncology

BACKGROUND

Efficient and accurate delineation of organs at risk (OARs) is a critical procedure for treatment planning and dose evaluation. Deep learning-based auto-segmentation of OARs has shown promising results and is increasingly being used in radiation therapy. However, existing deep learning-based auto-segmentation approaches face two challenges in clinical practice: generalizability and human-AI interaction. A generalizable and promptable auto-segmentation model, which segments OARs of multiple disease sites simultaneously and supports on-the-fly human-AI interaction, can significantly enhance the efficiency of radiation therapy treatment planning.

PURPOSE

Meta's segment anything model (SAM) was proposed as a generalizable and promptable model for next-generation natural image segmentation. We further evaluated the performance of SAM in radiotherapy segmentation.

METHODS

Computed tomography (CT) images of clinical cases from four disease sites at our institute were collected: prostate, lung, gastrointestinal, and head & neck. For each case, we selected the OARs important in radiotherapy treatment planning. We then compared both the Dice coefficients and Jaccard indices derived from three distinct methods: manual delineation (ground truth), automatic segmentation using SAM's 'segment anything' mode, and automatic segmentation using SAM's 'box prompt' mode that implements manual interaction via live prompts during segmentation.

RESULTS

Our results indicate that SAM's segment anything mode can achieve clinically acceptable segmentation results in most OARs with Dice scores higher than 0.7. SAM's box prompt mode further improves Dice scores by 0.1∼0.5. Similar results were observed for Jaccard indices. The results show that SAM performs better for prostate and lung, but worse for gastrointestinal and head & neck. When considering the size of organs and the distinctiveness of their boundaries, SAM shows better performance for large organs with distinct boundaries, such as lung and liver, and worse for smaller organs with less distinct boundaries, like parotid and cochlea.

CONCLUSIONS

Our results demonstrate SAM's robust generalizability with consistent accuracy in automatic segmentation for radiotherapy. Furthermore, the advanced box-prompt method enables the users to augment auto-segmentation interactively and dynamically, leading to patient-specific auto-segmentation in radiation therapy. SAM's generalizability across different disease sites and different modalities makes it feasible to develop a generic auto-segmentation model in radiotherapy.

Quality of Radiotherapy Workforce Training within the USA

The training, competency requirements and scope of practice of professionals within a radiation oncology department vary across countries. The purpose of this review is to shed light on the current status of radiotherapy training in the USA by discussing current benchmarks for medical residency, physics residency, radiation therapy and dosimetry training programmes. Although there are notable strengths, the US radiotherapy workforce training system also faces several challenges when it comes to standardising education to develop a competent workforce that meets societal needs. Continued efforts are needed at a systemic level to improve training in areas such as brachytherapy and proton therapy, promote research involvement and develop trainees who are equipped to form a competent radiation therapy workforce.

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.

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.

End-to-end deep learning radiomics: development and validation of a novel attention-based aggregate convolutional neural network to distinguish breast diffuse large B-cell lymphoma from breast invasive ductal carcinoma

Background

Apart from invasive pathological examination, there is no effective method to differentiate breast diffuse large B-cell lymphoma (DLBCL) from breast invasive ductal carcinoma (IDC). In this study, we aimed to develop and validate an effective deep learning radiomics model to discriminate between DLBCL and IDC.

Methods

A total of 324 breast nodules from 236 patients with baseline 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography/computed tomography (PET/CT) were retrospectively analyzed. After grouping breast DLBCL and breast IDC patients, external and internal datasets were divided according to the data collected by different centers. Preprocessing was then used to process the original PET/CT images and an attention-based aggregate convolutional neural network (AACNN) model was designed. The AACNN model was trained using patches of CT or PET tumor images and optimized with an improved loss function. The final ensemble predictive model was built using distance weight voting. Finally, the model performance was evaluated and statistically verified.

Results

A total of 249 breast nodules from Fudan University Shanghai Cancer Center (FUSCC) and 75 breast nodules from Shanghai Proton and Heavy Ion Center (SPHIC) were selected as internal and external datasets, respectively. On the internal testing, our method yielded an area under the curve (AUC), accuracy (ACC), sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV), and harmonic mean of precision and sensitivity (F1) of 0.886, 83.0%, 80.9%, 85.0%, 84.8%, 81.2%, and 0.828, respectively. Meanwhile on the external testing, the results were 0.788, 71.6%, 61.4%, 84.7%, 84.0%, 62.6%, and 0.709, respectively.

Conclusions

Our study outlines a deep learning radiomics method which can automatically, noninvasively, and accurately differentiate breast DLBCL from breast IDC, which will be more in line with the needs and strategies of precision medicine, individualized diagnosis, and treatment.

An investigation into the risk of population bias in deep learning autocontouring

BACKGROUND AND PURPOSE

To date, data used in the development of Deep Learning-based automatic contouring (DLC) algorithms have been largely sourced from single geographic populations. This study aimed to evaluate the risk of population-based bias by determining whether the performance of an autocontouring system is impacted by geographic population.

MATERIALS AND METHODS

80 Head Neck CT deidentified scans were collected from four clinics in Europe (n = 2) and Asia (n = 2). A single observer manually delineated 16 organs-at-risk in each. Subsequently, the data was contoured using a DLC solution, and trained using single institution (European) data. Autocontours were compared to manual delineations using quantitative measures. A Kruskal-Wallis test was used to test for any difference between populations. Clinical acceptability of automatic and manual contours to observers from each participating institution was assessed using a blinded subjective evaluation.

RESULTS

Seven organs showed a significant difference in volume between groups. Four organs showed statistical differences in quantitative similarity measures. The qualitative test showed greater variation in acceptance of contouring between observers than between data from different origins, with greater acceptance by the South Korean observers.

CONCLUSION

Much of the statistical difference in quantitative performance could be explained by the difference in organ volume impacting the contour similarity measures and the small sample size. However, the qualitative assessment suggests that observer perception bias has a greater impact on the apparent clinical acceptability than quantitatively observed differences. This investigation of potential geographic bias should extend to more patients, populations, and anatomical regions in the future.

Evaluating Capabilities of Large Language Models: Performance of GPT4 on Surgical Knowledge Assessments

Background

Artificial intelligence (AI) has the potential to dramatically alter healthcare by enhancing how we diagnosis and treat disease. One promising AI model is ChatGPT, a large general-purpose language model trained by OpenAI. The chat interface has shown robust, human-level performance on several professional and academic benchmarks. We sought to probe its performance and stability over time on surgical case questions.

Methods

We evaluated the performance of ChatGPT-4 on two surgical knowledge assessments: the Surgical Council on Resident Education (SCORE) and a second commonly used knowledge assessment, referred to as Data-B. Questions were entered in two formats: open-ended and multiple choice. ChatGPT output were assessed for accuracy and insights by surgeon evaluators. We categorized reasons for model errors and the stability of performance on repeat encounters.

Results

A total of 167 SCORE and 112 Data-B questions were presented to the ChatGPT interface. ChatGPT correctly answered 71% and 68% of multiple-choice SCORE and Data-B questions, respectively. For both open-ended and multiple-choice questions, approximately two-thirds of ChatGPT responses contained non-obvious insights. Common reasons for inaccurate responses included: inaccurate information in a complex question (n=16, 36.4%); inaccurate information in fact-based question (n=11, 25.0%); and accurate information with circumstantial discrepancy (n=6, 13.6%). Upon repeat query, the answer selected by ChatGPT varied for 36.4% of inaccurate questions; the response accuracy changed for 6/16 questions.

Conclusion

Consistent with prior findings, we demonstrate robust near or above human-level performance of ChatGPT within the surgical domain. Unique to this study, we demonstrate a substantial inconsistency in ChatGPT responses with repeat query. This finding warrants future consideration and presents an opportunity to further train these models to provide safe and consistent responses. Without mental and/or conceptual models, it is unclear whether language models such as ChatGPT would be able to safely assist clinicians in providing care.

DAART: a deep learning platform for deeply accelerated adaptive radiation therapy for lung cancer

Purpose

The study aimed to implement a novel, deeply accelerated adaptive radiation therapy (DAART) approach for lung cancer radiotherapy (RT). Lung cancer is the most common cause of cancer-related death, and RT is the preferred medically inoperable treatment for early stage non-small cell lung cancer (NSCLC). In the current lengthy workflow, it takes a median of four weeks from diagnosis to RT treatment, which can result in complete restaging and loss of local control with delay. We implemented the DAART approach, featuring a novel deepPERFECT system, to address unwanted delays between diagnosis and treatment initiation.

Materials and methods

We developed a deepPERFECT to adapt the initial diagnostic imaging to the treatment setup to allow initial RT planning and verification. We used data from 15 patients with NSCLC treated with RT to train the model and test its performance. We conducted a virtual clinical trial to evaluate the treatment quality of the proposed DAART for lung cancer radiotherapy.

Results

We found that deepPERFECT predicts planning CT with a mean high-intensity fidelity of 83 and 14 HU for the body and lungs, respectively. The shape of the body and lungs on the synthesized CT was highly conformal, with a dice similarity coefficient (DSC) of 0.91, 0.97, and Hausdorff distance (HD) of 7.9 mm, and 4.9 mm, respectively, compared with the planning CT scan. The tumor showed less conformality, which warrants acquisition of treatment Day1 CT and online adaptive RT. An initial plan was designed on synthesized CT and then adapted to treatment Day1 CT using the adapt to position (ATP) and adapt to shape (ATS) method. Non-inferior plan quality was achieved by the ATP scenario, while all ATS-adapted plans showed good plan quality.

Conclusion

DAART reduces the common online ART (ART) treatment course by at least two weeks, resulting in a 50% shorter time to treatment to lower the chance of restaging and loss of local control.

deepPERFECT: Novel Deep Learning CT Synthesis Method for Expeditious Pancreatic Cancer Radiotherapy

Major sources of delay in the standard of care RT workflow are the need for multiple appointments and separate image acquisition. In this work, we addressed the question of how we can expedite the workflow by synthesizing planning CT from diagnostic CT. This idea is based on the theory that diagnostic CT can be used for RT planning, but in practice, due to the differences in patient setup and acquisition techniques, separate planning CT is required. We developed a generative deep learning model, deepPERFECT, that is trained to capture these differences and generate deformation vector fields to transform diagnostic CT into preliminary planning CT. We performed detailed analysis both from an image quality and a dosimetric point of view, and showed that deepPERFECT enabled the preliminary RT planning to be used for preliminary and early plan dosimetric assessment and evaluation.

Fully automated imaging protocol independent system for pituitary adenoma segmentation: a convolutional neural network-based model on sparsely annotated MRI

This study aims to develop a fully automated imaging protocol independent system for pituitary adenoma segmentation from magnetic resonance imaging (MRI) scans that can work without user interaction and evaluate its accuracy and utility for clinical applications. We trained two independent artificial neural networks on MRI scans of 394 patients. The scans were acquired according to various imaging protocols over the course of 11 years on 1.5T and 3T MRI systems. The segmentation model assigned a class label to each input pixel (pituitary adenoma, internal carotid artery, normal pituitary gland, background). The slice segmentation model classified slices as clinically relevant (structures of interest in slice) or irrelevant (anterior or posterior to sella turcica). We used MRI data of another 99 patients to evaluate the performance of the model during training. We validated the model on a prospective cohort of 28 patients, Dice coefficients of 0.910, 0.719, and 0.240 for tumour, internal carotid artery, and normal gland labels, respectively, were achieved. The slice selection model achieved 82.5% accuracy, 88.7% sensitivity, 76.7% specificity, and an AUC of 0.904. A human expert rated 71.4% of the segmentation results as accurate, 21.4% as slightly inaccurate, and 7.1% as coarsely inaccurate. Our model achieved good results comparable with recent works of other authors on the largest dataset to date and generalized well for various imaging protocols. We discussed future clinical applications, and their considerations. Models and frameworks for clinical use have yet to be developed and evaluated.

Machine learning in non-small cell lung cancer radiotherapy: A bibliometric analysis

Background

Machine learning is now well-developed in non-small cell lung cancer (NSCLC) radiotherapy. But the research trend and hotspots are still unclear. To investigate the progress in machine learning in radiotherapy NSCLC, we performed a bibliometric analysis of associated research and discuss the current research hotspots and potential hot areas in the future.

Methods

The involved researches were obtained from the Web of Science Core Collection database (WoSCC). We used R-studio software, the Bibliometrix package and VOSviewer (Version 1.6.18) software to perform bibliometric analysis.

Results

We found 197 publications about machine learning in radiotherapy for NSCLC in the WoSCC, and the journal Medical Physics contributed the most articles. The University of Texas MD Anderson Cancer Center was the most frequent publishing institution, and the United States contributed most of the publications. In our bibliometric analysis, “radiomics” was the most frequent keyword, and we found that machine learning is mainly applied to analyze medical images in the radiotherapy of NSCLC.

Results

The research we identified about machine learning in NSCLC radiotherapy was mainly related to the radiotherapy planning of NSCLC and the prediction of treatment effects and adverse events in NSCLC patients who were under radiotherapy. Our research has added new insights into machine learning in NSCLC radiotherapy and could help researchers better identify hot research areas in the future.

Same-day adaptive palliative radiotherapy without prior CT simulation: Early outcomes in the FAST-METS study

BACKGROUND AND PURPOSE

Standard palliative radiotherapy workflows involve waiting times or multiple clinic visits. We developed and implemented a rapid palliative workflow using diagnostic imaging (dCT) for pre-planning, with subsequent on-couch target and plan adaptation based on a synthetic computed tomography (CT) obtained from cone-beam CT imaging (CBCT).

MATERIALS AND METHODS

Patients with painful bone metastases and recent diagnostic imaging were eligible for inclusion in this prospective, ethics-approved study. The workflow consisted of 1) telephone consultation with a radiation oncologist (RO); 2) pre-planning on the dCT using planning templates and mostly intensity-modulated radiotherapy; 3) RO consultation on the day of treatment; 4) CBCT scan with on-couch adaptation of the target and treatment plan; 5) delivery of either scheduled or adapted treatment plan. Primary outcomes were dosimetric data and treatment times; secondary outcome was patient satisfaction.

RESULTS

47 patients were enrolled between December 2021 and October 2022. In all treatments, adapted treatment plans were chosen due to significant improvements in target coverage (PTV/CTV V95%, p-value < 0.005) compared to the original treatment plan calculated on daily anatomy. Most patients were satisfied with the workflow. The average treatment time, including consultation and on-couch adaptive treatment, was 85 minutes. On-couch adaptation took on average 30 min. but was longer in cases where the automated deformable image registration failed to correctly propagate the targets.

CONCLUSION

A fast treatment workflow for patients referred for painful bone metastases was implemented successfully using online adaptive radiotherapy, without a dedicated CT simulation. Patients were generally satisfied with the palliative radiotherapy workflow.

DeSeg: auto detector-based segmentation for brain metastases

Delineation of brain metastases (BMs) is a paramount step in stereotactic radiosurgery treatment. Clinical practice has specific expectation on BM auto-delineation that the method is supposed to avoid missing of small lesions and yield accurate contours for large lesions. In this study, we propose a novel coarse-to-fine framework, named detector-based segmentation (DeSeg), to incorporate object-level detection into pixel-wise segmentation so as to meet the clinical demand. DeSeg consists of three components: a center-point-guided single-shot detector to localize the potential lesion regions, a multi-head U-Net segmentation model to refine contours, and a data cascade unit to connect both tasks smoothly. Performance on tiny lesions is measured by the object-based sensitivity and positive predictive value (PPV), while that on large lesions is quantified by dice similarity coefficient (DSC), average symmetric surface distance (ASSD) and 95% Hausdorff distance (HD95). Besides, computational complexity is also considered to study the potential of method in real-time processing. This study retrospectively collected 240 BM patients with Gadolinium injected contrast-enhanced T1-weighted magnetic resonance imaging (T1c-MRI), which were randomly split into training, validating and testing datasets (192, 24 and 24 scans, respectively). The lesions in the testing dataset were further divided into two groups based on the volume size (smallS: ≤1.5 cc,N= 88; largeL: > 1.5 cc,N= 15). On average, DeSeg yielded a sensitivity of 0.91 and a PPV of 0.77 on S group, and a DSC of 0.86, an ASSD 0f 0.76 mm and a HD95 of 2.31 mm onLgroup. The results indicated that DeSeg achieved leading sensitivity and PPV for tiny lesions as well as segmentation metrics for large ones. After our clinical validation, DeSeg showed competitive segmentation performance while kept faster processing speed comparing with existing 3D models.

Beyond high hopes: A scoping review of the 2019-2021 scientific discourse on machine learning in medical imaging

Machine learning has become a key driver of the digital health revolution. That comes with a fair share of high hopes and hype. We conducted a scoping review on machine learning in medical imaging, providing a comprehensive outlook of the field's potential, limitations, and future directions. Most reported strengths and promises included: improved (a) analytic power, (b) efficiency (c) decision making, and (d) equity. Most reported challenges included: (a) structural barriers and imaging heterogeneity, (b) scarcity of well-annotated, representative and interconnected imaging datasets (c) validity and performance limitations, including bias and equity issues, and (d) the still missing clinical integration. The boundaries between strengths and challenges, with cross-cutting ethical and regulatory implications, remain blurred. The literature emphasizes explainability and trustworthiness, with a largely missing discussion about the specific technical and regulatory challenges surrounding these concepts. Future trends are expected to shift towards multi-source models, combining imaging with an array of other data, in a more open access, and explainable manner.

The role of artificial intelligence in veterinary radiation oncology

Veterinary radiation oncology regularly deploys sophisticated contouring, image registration, and treatment planning optimization software for patient care. Over the past decade, advances in computing power and the rapid development of neural networks, open-source software packages, and data science have been realized and resulted in new research and clinical applications of artificial intelligent (AI) systems in radiation oncology. These technologies differ from conventional software in their level of complexity and ability to learn from representative and local data. We provide clinical and research application examples of AI in human radiation oncology and their potential applications in veterinary medicine throughout the patient's care-path: from treatment simulation, deformable registration, auto-segmentation, automated treatment planning and plan selection, quality assurance, adaptive radiotherapy, and outcomes modeling. These technologies have the potential to offer significant time and cost savings in the veterinary setting; however, since the range of usefulness of these technologies have not been well studied nor understood, care must be taken if adopting AI technologies in clinical practice. Over the next several years, some practical and realizable applications of AI in veterinary radiation oncology include automated segmentation of normal tissues and tumor volumes, deformable registration, multi-criteria plan optimization, and adaptive radiotherapy. Keys in achieving success in adopting AI in veterinary radiation oncology include: establishing "truth-data"; data harmonization; multi-institutional data and collaborations; standardized dose reporting and taxonomy; adopting an open access philosophy, data collection and curation; open-source algorithm development; and transparent and platform-independent code development.

Efficient dose-volume histogram-based pretreatment patient-specific quality assurance methodology with combined deep learning and machine learning models for volumetric modulated arc radiotherapy

BACKGROUND

Weak correlation between gamma passing rates and dose differences in target volumes and organs at risk (OARs) has been reported in several studies. Evaluation on the differences between planned dose-volume histogram (DVH) and reconstructed DVH from measurement was adopted and incorporated into patient-specific quality assurance (PSQA). However, it is difficult to develop a methodology allowing the evaluation of errors on DVHs accurately and quickly.

PURPOSE

To develop a DVH-based pretreatment PSQA for volumetric modulated arc therapy (VMAT) with combined deep learning (DL) and machine learning models to overcome the limitation of conventional gamma index (GI) and improve the efficiency of DVH-based PSQA.

METHODS

A DL model with a three-dimensional squeeze-and-excitation residual blocks incorporated into a modified U-net was developed to predict the measured PSQA DVHs of 208 head-and-neck (H&N) cancer patients underwent VMAT between 2018 and 2021 from two hospitals, in which 162 cases was randomly selected for training, 18 for validation, and 28 for testing. After evaluating the differences between treatment planning system (TPS) and PSQA DVHs predicted by DL model with multiple metrics, a pass or fail (PoF) classification model was developed using XGBoost algorithm. Evaluation of domain experts on dose errors between TPS and reconstructed PSQA DVHs was taken as ground truth for PoF classification model training.

RESULTS

The prediction model was able to achieve a good agreement between predicted, measured, and TPS doses. Quantitative evaluation demonstrated no significant difference between predicted PSQA dose and measured dose for target and OARs, except for D_mean of PTV6900 (p = 0.001), D₅₀ of PTV6000 (p = 0.014), D₂ of PTV5400 (p = 0.009), D₅₀ of left parotid (p = 0.015), and D_max of left inner ear (p = 0.007). The XGBoost model achieved an area under curves, accuracy, sensitivity, and specificity of 0.89 versus 0.88, 0.89 versus 0.86, 0. 71 versus 0.71, and 0.95 versus 0.91 with measured and predicted PSQA doses, respectively. The agreement between domain experts and the classification model was 86% for 28 test cases.

CONCLUSIONS

The successful prediction of PSQA doses and classification of PoF for H&N VMAT PSQA indicating that this DVH-based PSQA method is promising to overcome the limitations of GI and to improve the efficiency and accuracy of VMAT delivery.

Perceptions of Canadian vascular surgeons toward artificial intelligence and machine learning

Background

Artificial intelligence (AI) and machine learning (ML) are rapidly advancing fields with increasing utility in health care. We conducted a survey to determine the perceptions of Canadian vascular surgeons toward AI/ML.

Methods

An online questionnaire was distributed to 162 members of the Canadian Society for Vascular Surgery. Self-reported knowledge, attitudes, and perceptions with respect to potential applications, limitations, and facilitators of AI/ML were assessed.

Results

Overall, 50 of the 162 Canadian vascular surgeons (31%) responded to the survey. Most respondents were aged 30 to 59 years (72%), male (80%), and White (67%) and practiced in academic settings (72%). One half of the participants reported that their knowledge of AI/ML was poor or very poor. Most were excited or very excited about AI/ML (66%) and were interested or very interested in learning more about the field (83.7%). The respondents believed that AI/ML would be useful or very useful for diagnosis (62%), prognosis (72%), patient selection (56%), image analysis (64%), intraoperative guidance (52%), research (88%), and education (80%). The limitations that the participants were most concerned about were errors leading to patient harm (42%), bias based on patient demographics (42%), and lack of clinician knowledge and skills in AI/ML (40%). Most were not concerned or were mildly concerned about job replacement (86%). The factors that were most important to encouraging clinicians to use AI/ML models were improvements in efficiency (88%), accurate predictions (84%), and ease of use (84%). The comments from respondents focused on the pressing need for the implementation of AI/ML in vascular surgery owing to the potential to improve care delivery.

Conclusions

Canadian vascular surgeons have positive views on AI/ML and believe this technology can be applied to multiple aspects of the specialty to improve patient care, research, and education. Current self-reported knowledge is poor, although interest was expressed in learning more about the field. The facilitators and barriers to the effective use of AI/ML identified in the present study can guide future development of these tools in vascular surgery.

A Systematic Review of Polygenic Models for Predicting Drug Outcomes

Polygenic models have emerged as promising prediction tools for the prediction of complex traits. Currently, the majority of polygenic models are developed in the context of predicting disease risk, but polygenic models may also prove useful in predicting drug outcomes. This study sought to understand how polygenic models incorporating pharmacogenetic variants are being used in the prediction of drug outcomes. A systematic review was conducted with the aim of gaining insights into the methods used to construct polygenic models, as well as their performance in drug outcome prediction. The search uncovered 89 papers that incorporated pharmacogenetic variants in the development of polygenic models. It was found that the most common polygenic models were constructed for drug dosing predictions in anticoagulant therapies (n = 27). While nearly all studies found a significant association with their polygenic model and the investigated drug outcome (93.3%), less than half (47.2%) compared the performance of the polygenic model against clinical predictors, and even fewer (40.4%) sought to validate model predictions in an independent cohort. Additionally, the heterogeneity of reported performance measures makes the comparison of models across studies challenging. These findings highlight key considerations for future work in developing polygenic models in pharmacogenomic research.

Fully automatic prognostic biomarker extraction from metastatic prostate lesion segmentations in whole-body [68Ga]Ga-PSMA-11 PET/CT images

PURPOSE

This study aimed to develop and assess an automated segmentation framework based on deep learning for metastatic prostate cancer (mPCa) lesions in whole-body [⁶⁸Ga]Ga-PSMA-11 PET/CT images for the purpose of extracting patient-level prognostic biomarkers.

METHODS

Three hundred thirty-seven [⁶⁸Ga]Ga-PSMA-11 PET/CT images were retrieved from a cohort of biochemically recurrent PCa patients. A fully 3D convolutional neural network (CNN) is proposed which is based on the self-configuring nnU-Net framework, and was trained on a subset of these scans, with an independent test set reserved for model evaluation. Voxel-level segmentation results were assessed using the dice similarity coefficient (DSC), positive predictive value (PPV), and sensitivity. Sensitivity and PPV were calculated to assess lesion level detection; patient-level classification results were assessed by the accuracy, PPV, and sensitivity. Whole-body biomarkers total lesional volume (TLV_auto) and total lesional uptake (TLU_auto) were calculated from the automated segmentations, and Kaplan-Meier analysis was used to assess biomarker relationship with patient overall survival.

RESULTS

At the patient level, the accuracy, sensitivity, and PPV were all > 90%, with the best metric being the PPV (97.2%). PPV and sensitivity at the lesion level were 88.2% and 73.0%, respectively. DSC and PPV measured at the voxel level performed within measured inter-observer variability (DSC, median = 50.7% vs. second observer = 32%, p = 0.012; PPV, median = 64.9% vs. second observer = 25.7%, p < 0.005). Kaplan-Meier analysis of TLV_auto and TLU_auto showed they were significantly associated with patient overall survival (both p < 0.005).

CONCLUSION

The fully automated assessment of whole-body [⁶⁸Ga]Ga-PSMA-11 PET/CT images using deep learning shows significant promise, yielding accurate scan classification, voxel-level segmentations within inter-observer variability, and potentially clinically useful prognostic biomarkers associated with patient overall survival.

TRIAL REGISTRATION

This study was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12615000608561) on 11 June 2015.

Reflections on beam configuration optimization for intensity-modulated proton therapy

Presumably, intensity-modulated proton radiotherapy (IMPT) is the most powerful form of proton radiotherapy. In the current state of the art, IMPT beam configurations (i.e. the number of beams and their directions) are, in general, chosen subjectively based on prior experience and practicality. Beam configuration optimization (BCO) for IMPT could, in theory, significantly enhance IMPT’s therapeutic potential. However, BCO is complex and highly computer resource-intensive. Some algorithms for BCO have been developed for intensity-modulated photon therapy (IMRT). They are rarely used clinically mainly because the large number of beams typically employed in IMRT renders BCO essentially unnecessary. Moreover, in the newer form of IMRT, volumetric modulated arc therapy, there are no individual static beams. BCO is of greater importance for IMPT because it typically employs a very small number of beams (2-4) and, when the number of beams is small, BCO is critical for improving plan quality. However, the unique properties and requirements of protons, particularly in IMPT, make BCO challenging. Protons are more sensitive than photons to anatomic changes, exhibit variable relative biological effectiveness along their paths, and, as recently discovered, may spare the immune system. Such factors must be considered in IMPT BCO, though doing so would make BCO more resource intensive and make it more challenging to extend BCO algorithms developed for IMRT to IMPT. A limited amount of research in IMPT BCO has been conducted; however, considerable additional work is needed for its further development to make it truly effective and computationally practical. This article aims to provide a review of existing BCO algorithms, most of which were developed for IMRT, and addresses important requirements specific to BCO for IMPT optimization that necessitate the modification of existing approaches or the development of new effective and efficient ones.

Beyond automatic medical image segmentation-the spectrum between fully manual and fully automatic delineation

Semi-automatic and fully automatic contouring tools have emerged as an alternative to fully manual segmentation to reduce time spent contouring and to increase contour quality and consistency. Particularly, fully automatic segmentation has seen exceptional improvements through the use of deep learning in recent years. These fully automatic methods may not require user interactions, but the resulting contours are often not suitable to be used in clinical practice without a review by the clinician. Furthermore, they need large amounts of labelled data to be available for training. This review presents alternatives to manual or fully automatic segmentation methods along the spectrum of variable user interactivity and data availability. The challenge lies to determine how much user interaction is necessary and how this user interaction can be used most effectively. While deep learning is already widely used for fully automatic tools, interactive methods are just at the starting point to be transformed by it. Interaction between clinician and machine, via artificial intelligence, can go both ways and this review will present the avenues that are being pursued to improve medical image segmentation.

Machine learning for contour classification in TG-263 noncompliant databases

A large volume of medical data are labeled using nonstandardized nomenclature. Although efforts have been made by the American Association of Physicists in Medicine (AAPM) to standardize nomenclature through Task Group 263 (TG-263), there remain noncompliant databases. This work aims to create an algorithm that can analyze anatomical contours in patients with head and neck cancer and classify them into TG-263 compliant nomenclature. To create an accurate algorithm capable of such classification, a combined approaching using both binary images of individual slices of anatomical contours themselves, as well as center of mass coordinates of the structures are input into a neural network. The center of mass coordinates were scaled using two normalization schemes, a simple linear normalization scheme agnostic of the patient anatomy, and an anatomical normalization scheme dependent on patient anatomy. The results of all of the individual slice classifications are then aggregated into a single classification by means of a voting algorithm. The total classification accuracy of the final algorithms was 97.6% mean accuracy per class for nonanatomically normalization scheme, and 97.9% mean accuracy per class for anatomically normalization scheme. The total accuracy was 99.0% (13 errors in 1302 structures) for the nonanatomically normalization scheme, and 98.3% (22 errors in 1302 structures) for the anatomically normalization scheme.

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.

Barriers and facilitators to clinical implementation of radiotherapy treatment planning automation: A survey study of medical dosimetrists

PURPOSE

Little is known about the scale of clinical implementation of automated treatment planning techniques in the United States. In this work, we examine the barriers and facilitators to adoption of commercially available automated planning tools into the clinical workflow using a survey of medical dosimetrists.

METHODS/MATERIALS

Survey questions were developed based on a literature review of automation research and cognitive interviews of medical dosimetrists at our institution. Treatment planning automation was defined to include auto-contouring and automated treatment planning. Survey questions probed frequency of use, positive and negative perceptions, potential implementation changes, and demographic and institutional descriptive statistics. The survey sample was identified using both a LinkedIn search and referral requests sent to physics directors and senior physicists at 34 radiotherapy clinics in our state. The survey was active from August 2020 to April 2021.

RESULTS

Thirty-four responses were collected out of 59 surveys sent. Three categories of barriers to use of automation were identified. The first related to perceptions of limited accuracy and usability of the algorithms. Eighty-eight percent of respondents reported that auto-contouring inaccuracy limited its use, and 62% thought it was difficult to modify an automated plan, thus limiting its usefulness. The second barrier relates to the perception that automation increases the probability of an error reaching the patient. Third, respondents were concerned that automation will make their jobs less satisfying and less secure. Large majorities reported that they enjoyed plan optimization, would not want to lose that part of their job, and expressed explicit job security fears.

CONCLUSION

To our knowledge this is the first systematic investigation into the views of automation by medical dosimetrists. Potential barriers and facilitators to use were explicitly identified. This investigation highlights several concrete approaches that could potentially increase the translation of automation into the clinic, along with areas of needed research.

Radiotherapy Standardisation and Artificial Intelligence within the National Cancer Institute's Clinical Trials Network

Artificial intelligence in healthcare refers to the use of complex algorithms designed to conduct certain tasks in an automated manner. Artificial intelligence has a transformative power in radiation oncology to improve the quality and efficiency of patient care, given the increase in volume and complexity of digital data, as well as the multi-faceted and highly technical nature of this field of medicine. However, artificial intelligence alone will not be able to fix healthcare's problem, because new technologies bring unexpected and potentially underappreciated obstacles. The inclusion of multicentre datasets, the incorporation of time-varying data, the assessment of missing data as well as informative censoring and the addition of clinical utility could significantly improve artificial intelligence models. Standardisation plays a crucial, supportive and leading role in artificial intelligence. Clinical trials are the most reliable method of demonstrating the efficacy and safety of a treatment or clinical approach, as well as providing high-level evidence to justify artificial intelligence. The National Surgical Adjuvant Breast and Bowel Project, the Radiation Therapy Oncology Group and the Gynecologic Oncology Group collaborated to form NRG Oncology (acronym NRG derived from the names of the parental groups). NRG Oncology is one of the adult cancer clinical trial groups containing radiotherapy specialty of the National Cancer Institute's Clinical Trials Network (NCTN). Standardisation from NRG/NCTN has the potential to reduce variation in clinical treatment and patient outcome by eliminating potential errors, enabling broader application of artificial intelligence tools. NCTN, NRG and Imaging and Radiation Oncology Core are in a unique position to help with standards development, advocacy and enforcement, all of which can benefit from artificial intelligence, as artificial intelligence has the ability to improve trial success rates by transforming crucial phases in clinical trial design, from study planning through to execution. Here we will examine: (i) how to conduct technical and clinical evaluations before adopting artificial intelligence technologies, (ii) how to obtain high-quality data for artificial intelligence, (iii) the NCTN infrastructure and standards, (iv) radiotherapy standardisation for clinical trials and (v) artificial intelligence applications in standardisation.

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.

Few-shot learning for deformable image registration in 4DCT images

OBJECTIVES

To develop a rapid and accurate 4D deformable image registration (DIR) approach for online adaptive radiotherapy.

METHODS

We propose a deep learning (DL)-based few-shot registration network (FR-Net) to generate deformation vector fields from each respiratory phase to an implicit reference image, thereby mitigating the bias introduced by the selection of reference images. The proposed FR-Net is pretrained with limited unlabeled 4D data and further optimized by maximizing the intensity similarity of one specific four-dimensional computed tomography (4DCT) scan. Because of the learning ability of DL models, the few-shot learning strategy facilitates the generalization of the model to other 4D data sets and the acceleration of the optimization process.

RESULTS

The proposed FR-Net is evaluated for 4D groupwise and 3D pairwise registration on thoracic 4DCT data sets DIR-Lab and POPI. FR-Net displays an averaged target registration error of 1.48 mm and 1.16 mm between the maximum inhalation and exhalation phases in the 4DCT of DIR-Lab and POPI, respectively, with approximately 2 min required to optimize one 4DCT. Overall, FR-Net outperforms state-of-the-art methods in terms of registration accuracy and exhibits a low computational time.

CONCLUSION

We develop a few-shot groupwise DIR algorithm for 4DCT images. The promising registration performance and computational efficiency demonstrate the prospective applications of this approach in registration tasks for online adaptive radiotherapy.

ADVANCES IN KNOWLEDGE

This work exploits DL models to solve the optimization problem in registering 4DCT scans while combining groupwise registration and few-shot learning strategy to solve the problem of consuming computational time and inferior registration accuracy.

Varian ethos online adaptive radiotherapy for prostate cancer: Early results of contouring accuracy, treatment plan quality, and treatment time

The Varian Ethos system allows for online adaptive treatments through the utilization of artificial intelligence (AI) and deformable image registration which automates large parts of the anatomical contouring and plan optimization process. In this study, treatments of intact prostate and prostate bed, with and without nodes, were simulated for 182 online adaptive fractions, and then a further 184 clinical fractions were delivered on the Ethos system. Frequency and magnitude of contour edits were recorded, as well as a range of plan quality metrics. From the fractions analyzed, 11% of AI generated contours, known as influencer contours, required no change, and 81% required minor edits in any given fraction. The frequency of target and noninfluencer organs at risk (OAR) contour editing varied substantially between different targets and noninfluencer OARs, although across all targets 72% of cases required no edits. The adaptive plan was the preference in 95% of fractions. The adaptive plan met more goals than the scheduled plan in 78% of fractions, while in 15% of fractions the number of goals met was the same. The online adaptive recontouring and replanning process was carried out in 19 min on average. Significant improvements in dosimetry are possible with the Ethos online adaptive system in prostate radiotherapy.

Enhancing the impact of Artificial Intelligence in Medicine: A joint AIFM-INFN Italian initiative for a dedicated cloud-based computing infrastructure

Artificial Intelligence (AI) techniques have been implemented in the field of Medical Imaging for more than forty years. Medical Physicists, Clinicians and Computer Scientists have been collaborating since the beginning to realize software solutions to enhance the informative content of medical images, including AI-based support systems for image interpretation. Despite the recent massive progress in this field due to the current emphasis on Radiomics, Machine Learning and Deep Learning, there are still some barriers to overcome before these tools are fully integrated into the clinical workflows to finally enable a precision medicine approach to patients' care. Nowadays, as Medical Imaging has entered the Big Data era, innovative solutions to efficiently deal with huge amounts of data and to exploit large and distributed computing resources are urgently needed. In the framework of a collaboration agreement between the Italian Association of Medical Physicists (AIFM) and the National Institute for Nuclear Physics (INFN), we propose a model of an intensive computing infrastructure, especially suited for training AI models, equipped with secure storage systems, compliant with data protection regulation, which will accelerate the development and extensive validation of AI-based solutions in the Medical Imaging field of research. This solution can be developed and made operational by Physicists and Computer Scientists working on complementary fields of research in Physics, such as High Energy Physics and Medical Physics, who have all the necessary skills to tailor the AI-technology to the needs of the Medical Imaging community and to shorten the pathway towards the clinical applicability of AI-based decision support systems.

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.

Intentional deep overfit learning (IDOL): A novel deep learning strategy for adaptive radiation therapy

PURPOSE

Applications of deep learning (DL) are essential to realizing an effective adaptive radiotherapy (ART) workflow. Despite the promise demonstrated by DL approaches in several critical ART tasks, there remain unsolved challenges to achieve satisfactory generalizability of a trained model in a clinical setting. Foremost among these is the difficulty of collecting a task-specific training dataset with high-quality, consistent annotations for supervised learning applications. In this study, we propose a tailored DL framework for patient-specific performance that leverages the behavior of a model intentionally overfitted to a patient-specific training dataset augmented from the prior information available in an ART workflow-an approach we term Intentional Deep Overfit Learning (IDOL).

METHODS

Implementing the IDOL framework in any task in radiotherapy consists of two training stages: (1) training a generalized model with a diverse training dataset of N patients, just as in the conventional DL approach, and (2) intentionally overfitting this general model to a small training dataset-specific the patient of interest ( N + 1 ) generated through perturbations and augmentations of the available task- and patient-specific prior information to establish a personalized IDOL model. The IDOL framework itself is task-agnostic and is, thus, widely applicable to many components of the ART workflow, three of which we use as a proof of concept here: the autocontouring task on replanning CTs for traditional ART, the MRI super-resolution (SR) task for MRI-guided ART, and the synthetic CT (sCT) reconstruction task for MRI-only ART.

RESULTS

In the replanning CT autocontouring task, the accuracy measured by the Dice similarity coefficient improves from 0.847 with the general model to 0.935 by adopting the IDOL model. In the case of MRI SR, the mean absolute error (MAE) is improved by 40% using the IDOL framework over the conventional model. Finally, in the sCT reconstruction task, the MAE is reduced from 68 to 22 HU by utilizing the IDOL framework.

CONCLUSIONS

In this study, we propose a novel IDOL framework for ART and demonstrate its feasibility using three ART tasks. We expect the IDOL framework to be especially useful in creating personally tailored models in situations with limited availability of training data but existing prior information, which is usually true in the medical setting in general and is especially true in ART.

Implementation of web-based open-source radiotherapy delineation software (WORDS) in organs at risk contouring training for newly qualified radiotherapists: quantitative comparison with conventional one-to-one coaching approach

BACKGROUND

Due to the role expansion of radiotherapists in dosimetric aspect, radiotherapists have taken up organs at risk (OARs) contouring work in many clinical settings. However, training of newly qualified radiotherapists in OARs contouring can be time consuming, it may also cause extra burden to experienced radiotherapists. As web-based open-source radiotherapy delineation software (WORDS) has become more readily available, it has provided a free and interactive alternative to conventional one-to-one coaching approach during OARs contouring training. The present study aims to evaluate the effectiveness of WORDS in training OARs contouring skills of newly qualified radiotherapists, compared to those trained by conventional one-to-one coaching approach.

METHODS

Nine newly qualified radiotherapists (licensed in 2017 - 2018) were enrolled to the conventional one-to-one coaching group (control group), while 11 newly qualified radiotherapists (licensed in 2019 - 2021) were assigned to WORDS training group (measured group). Ten OARs were selected to be contoured in this 3-phases quantitative study. Participants were required to undergo phase 1 OARs contouring in the beginning of the training session. Afterwards, conventional one-to-one training or WORDS training session was provided to participants according to their assigned group. Then the participants did phase 2 and 3 OARs contouring which were separated 1 week apart. Phase 1 - 3 OARs contouring aimed to demonstrate participants' pre-training OARs contouring ability, post-training OARs contouring ability and knowledge retention after one-week interval respectively using either training approach. To prevent bias, the computed tomography dataset for OARs contouring in each phase were different. Variations in the contouring scores for the selected OARs were evaluated between 3 phases using Kruskal-Wallis tests with Dunn tests for pairwise comparisons. Variations in the contouring scores between control and measured group in phase 1 - 3 contouring were analyzed using Wilcoxon signed-rank test. A p-value < 0.05 was considered to be statistically significant.

RESULTS

In both control group and measured group, significant improvement (p < 0.05) in phase 2 and 3 contouring scores have been observed comparing to phase 1 contouring scores. In comparison of contouring scores between control group and measured group, no significant differences (p > 0.05) were observed in all OARs between both groups.

CONCLUSIONS

The results in this study have demonstrated that the outcome of OARs contouring training using WORDS is comparable to the conventional training approach. In addition, WORDS can offer flexibility to newly qualified radiotherapists to practice OARs contouring at will, as well as reduce staff training burden of experienced radiotherapists.

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.

Radiographers' perspectives on the emerging integration of artificial intelligence into diagnostic imaging: The Ghana study

INTRODUCTION

The integration of artificial intelligence (AI) systems into medical imaging is advancing the practice and patient care. It is thought to further revolutionise the entire field in the near future. This study explored Ghanaian radiographers' perspectives on the integration of AI into medical imaging.

METHODS

A cross-sectional online survey of registered Ghanaian radiographers was conducted within a 3-month period (February-April, 2020). The survey sought information relating to demography, general perspectives on AI and implementation issues. Descriptive and inferential statistics were used for data analyses.

RESULTS

A response rate of 64.5% (151/234) was achieved. Majority of the respondents (n = 122, 80.8%) agreed that AI technology is the future of medical imaging. A good number of them (n = 131, 87.4%) indicated that AI would have an overall positive impact on medical imaging practice. However, some expressed fears about AI-related errors (n = 126, 83.4%), while others expressed concerns relating to job security (n = 35, 23.2%). High equipment cost, lack of knowledge and fear of cyber threats were identified as some factors hindering AI implementation in Ghana.

CONCLUSIONS

The radiographers who responded to this survey demonstrated a positive attitude towards the integration of AI into medical imaging. However, there were concerns about AI-related errors, job displacement and salary reduction which need to be addressed. Lack of knowledge, high equipment cost and cyber threats could impede the implementation of AI in medical imaging in Ghana. These findings are likely comparable to most low resource countries and we suggest more education to promote credibility of AI in practice.

Artificial intelligence: The opinions of radiographers and radiation therapists in Ireland

INTRODUCTION

Implementation of Artificial Intelligence (AI) into medical imaging is much debated. Diagnostic Radiographers (DRs) and Radiation Therapists (RTTs) are at the forefront of this technological leap, thus an understanding of their views, in particular changes to their current roles, is key to safe, optimal implementation.

METHODS

An online survey was designed, including themes: role changes, clinical priorities for AI, patient benefits, and education. It was distributed nationally in the Republic of Ireland via the national professional body, clinical management, and social media.

RESULTS

318 DRs and 77 RTTs participated. Priority areas for development included quality assurance, clinical audit, radiation dose optimisation, and improved workflow for DRs and treatment planning algorithm optimisation, clinical audit, and post processing for RTTs. There was resistance regarding AI use for patient facing roles and final image interpretation. 27.6% of DRs and 40.3% of RTTs currently use AI clinically and 46.1% of DRs and 41.2% of RTTs anticipate reduced staffing levels with AI. 64.9% of DRs and 70.6% of RTTs felt AI will be positive for patients, with the majority promoting AI regulation through national legislation. 86.1% of DRs and 94.0% of RTTs were favourable to AI implementation.

CONCLUSION

This research identifies priority AI development and implementation areas for DRs and RTTs. It thus highlights that DRs and RTTs should be involved in development of AI tools that would best support practice, and that clearly defined pathways for AI implementation into these key professions requires discussion so that optimum use and patient safety can ensue.

IMPLICATIONS FOR PRACTICE

Understanding opinions of AI has significant implications for practice, for ensuring optimal product development, implementation, and training, together with planning for potential DR and RTT role changes.

Deep learning based synthetic-CT generation in radiotherapy and PET: A review

Recently,deep learning (DL)-based methods for the generation of synthetic computed tomography (sCT) have received significant research attention as an alternative to classical ones. We present here a systematic review of these methods by grouping them into three categories, according to their clinical applications: (i) to replace computed tomography in magnetic resonance (MR) based treatment planning, (ii) facilitate cone-beam computed tomography based image-guided adaptive radiotherapy, and (iii) derive attenuation maps for the correction of positron emission tomography. Appropriate database searching was performed on journal articles published between January 2014 and December 2020. The DL methods' key characteristics were extracted from each eligible study, and a comprehensive comparison among network architectures and metrics was reported. A detailed review of each category was given, highlighting essential contributions, identifying specific challenges, and summarizing the achievements. Lastly, the statistics of all the cited works from various aspects were analyzed, revealing the popularity and future trends and the potential of DL-based sCT generation. The current status of DL-based sCT generation was evaluated, assessing the clinical readiness of the presented methods.