Deep learning for high-resolution dose prediction in high dose rate brachytherapy for breast cancer treatment.

Objective.Monte Carlo (MC) simulations are the benchmark for accurate radiotherapy dose calculations, notably in patient-specific high dose rate brachytherapy (HDR BT), in cases where considering tissue heterogeneities is critical. However, the lengthy computational time limits the practical application of MC simulations. Prior research used deep learning (DL) for dose prediction as an alternative to MC simulations. While accurate dose predictions akin to MC were attained, graphics processing unit limitations constrained these predictions to large voxels of 3 mm × 3 mm × 3 mm. This study aimed to enable dose predictions as accurate as MC simulations in 1 mm × 1 mm × 1 mm voxels within a clinically acceptable timeframe.Approach.Computed tomography scans of 98 breast cancer patients treated with Iridium-192-based HDR BT were used: 70 for training, 14 for validation, and 14 for testing. A new cropping strategy based on the distance to the seed was devised to reduce the volume size, enabling efficient training of 3D DL models using 1 mm × 1 mm × 1 mm dose grids. Additionally, novel DL architecture with layer-level fusion were proposed to predict MC simulated dose to medium-in-medium (Dm,m). These architectures fuse information from TG-43 dose to water-in-water (Dw,w) with patient tissue composition at the layer-level. Different inputs describing patient body composition were investigated.Main results.The proposed approach demonstrated state-of-the-art performance, on par with the MCDm,mmaps, but 300 times faster. The mean absolute percent error for dosimetric indices between the MC and DL-predicted complete treatment plans was 0.17% ± 0.15% for the planning target volumeV100, 0.30% ± 0.32% for the skinD2cc, 0.82% ± 0.79% for the lungD2cc, 0.34% ± 0.29% for the chest wallD2ccand 1.08% ± 0.98% for the heartD2cc.Significance.Unlike the time-consuming MC simulations, the proposed novel strategy efficiently converts TG-43Dw,wmaps into preciseDm,mmaps at high resolution, enabling clinical integration.

Penalty weight tuning in high dose rate brachytherapy using multi-objective Bayesian optimization.

OBJECTIVE: Treatment plan optimization in high dose rate (HDR) brachytherapy often requires manual fine-tuning of penalty weights for each objective, which can be time-consuming and dependent on the planner's experience. To automate this process, this study used a multi-criteria approach called multi-objective Bayesian optimization with q-noisy expected hypervolume improvement as its acquisition function (MOBO-qNEHVI). Approach: The treatment plans of 13 prostate cancer patients were retrospectively imported to a research treatment planning system, RapidBrachyMTPS, where fast mixed integer optimization (FMIO) performs dwell time optimization given a set of penalty weights to deliver 15 Gy to the target volume. MOBO-qNEHVI was used to find patient-specific Pareto optimal penalty weight vectors that yield clinically acceptable dose volume histogram metrics. The relationship between the number of MOBO-qNEHVI iterations and the number of clinically acceptable plans per patient (acceptance rate) was investigated. The performance time was obtained for various parameter configurations. Main results: MOBO-qNEHVI found clinically acceptable treatment plans for all patients. With increasing the number of MOBO-qNEHVI iterations, the acceptance rate grew logarithmically while the performance time grew exponentially. Fixing the penalty weight of the tumour volume to maximum value, adding the target dose as a parameter, initiating MOBO-qNEHVI with 25 parallel sampling of FMIO, and running 6 MOBO-qNEHVI iterations found solutions that delivered 15 Gy to the hottest 95% of the clinical target volume while respecting the dose constraints to the organs at risk. The average acceptance rate for each patient was 89.74% ± 8.11%, and performance time was 66.6 ± 12.6 seconds. The initiation took 22.47 ± 7.57 s, and each iteration took 7.35 ± 2.45 s to find one Pareto solution. Significance: MOBO-qNEHVI can automatically explore the trade-offs between treatment plan objectives in a patient-specific manner within a minute. This approach can reduce the dependency of plan quality on planner's experience.

A Deep Learning-Based Approach to Estimate Paneth Cell Granule Area in Celiac Disease.

CONTEXT.­: Changes in Paneth cell numbers can be associated with chronic inflammatory diseases of the gastrointestinal tract. So far, no consensus has been achieved on the number of Paneth cells and their relevance to celiac disease (CD). OBJECTIVES.­: To compare crypt and Paneth cell granule areas between patients with CD and without CD (non-CD) using an artificial intelligence-based solution. DESIGN.­: Hematoxylin-eosin-stained sections of duodenal biopsies from 349 patients at the McGill University Health Centre were analyzed. Of these, 185 had a history of CD and 164 were controls. Slides were digitized and NoCodeSeg, a code-free workflow using open-source software (QuPath, DeepMIB), was implemented to train deep learning models to segment crypts and Paneth cell granules. The total area of the entire analyzed tissue, epithelium, crypts, and Paneth cell granules was documented for all slides, and comparisons were performed. RESULTS.­: A mean intersection-over-union score of 88.76% and 91.30% was achieved for crypt areas and Paneth cell granule segmentations, respectively. On normalization to total tissue area, the crypt to total tissue area in CD was increased and Paneth cell granule area to total tissue area decreased when compared to non-CD controls. CONCLUSIONS.­: Crypt hyperplasia was confirmed in CD compared to non-CD controls. The area of Paneth cell granules, an indirect measure of Paneth cell function, decreased with increasing severity of CD. More importantly, our study analyzed complete hematoxylin-eosin slide sections using an efficient and easy to use coding-free artificial intelligence workflow.

Semi-Self-Supervised Learning for Semantic Segmentation in Images with Dense Patterns.

Deep learning has shown potential in domains with large-scale annotated datasets. However, manual annotation is expensive, time-consuming, and tedious. Pixel-level annotations are particularly costly for semantic segmentation in images with dense irregular patterns of object instances, such as in plant images. In this work, we propose a method for developing high-performing deep learning models for semantic segmentation of such images utilizing little manual annotation. As a use case, we focus on wheat head segmentation. We synthesize a computationally annotated dataset-using a few annotated images, a short unannotated video clip of a wheat field, and several video clips with no wheat-to train a customized U-Net model. Considering the distribution shift between the synthesized and real images, we apply three domain adaptation steps to gradually bridge the domain gap. Only using two annotated images, we achieved a Dice score of 0.89 on the internal test set. When further evaluated on a diverse external dataset collected from 18 different domains across five countries, this model achieved a Dice score of 0.73. To expose the model to images from different growth stages and environmental conditions, we incorporated two annotated images from each of the 18 domains to further fine-tune the model. This increased the Dice score to 0.91. The result highlights the utility of the proposed approach in the absence of large-annotated datasets. Although our use case is wheat head segmentation, the proposed approach can be extended to other segmentation tasks with similar characteristics of irregularly repeating patterns of object instances.

Generalizability of Machine Learning Models: Quantitative Evaluation of Three Methodological Pitfalls.

Purpose: To investigate the impact of the following three methodological pitfalls on model generalizability: (a) violation of the independence assumption, (b) model evaluation with an inappropriate performance indicator or baseline for comparison, and (c) batch effect. Materials and Methods: The authors used retrospective CT, histopathologic analysis, and radiography datasets to develop machine learning models with and without the three methodological pitfalls to quantitatively illustrate their effect on model performance and generalizability. F1 score was used to measure performance, and differences in performance between models developed with and without errors were assessed using the Wilcoxon rank sum test when applicable. Results: Violation of the independence assumption by applying oversampling, feature selection, and data augmentation before splitting data into training, validation, and test sets seemingly improved model F1 scores by 71.2% for predicting local recurrence and 5.0% for predicting 3-year overall survival in head and neck cancer and by 46.0% for distinguishing histopathologic patterns in lung cancer. Randomly distributing data points for a patient across datasets superficially improved the F1 score by 21.8%. High model performance metrics did not indicate high-quality lung segmentation. In the presence of a batch effect, a model built for pneumonia detection had an F1 score of 98.7% but correctly classified only 3.86% of samples from a new dataset of healthy patients. Conclusion: Machine learning models developed with these methodological pitfalls, which are undetectable during internal evaluation, produce inaccurate predictions; thus, understanding and avoiding these pitfalls is necessary for developing generalizable models.Keywords: Random Forest, Diagnosis, Prognosis, Convolutional Neural Network (CNN), Medical Image Analysis, Generalizability, Machine Learning, Deep Learning, Model Evaluation Supplemental material is available for this article. Published under a CC BY 4.0 license.

Radiomics as an emerging tool in the management of brain metastases.

Brain metastases (BM) are associated with significant morbidity and mortality in patients with advanced cancer. Despite significant advances in surgical, radiation, and systemic therapy in recent years, the median overall survival of patients with BM is less than 1 year. The acquisition of medical images, such as computed tomography (CT) and magnetic resonance imaging (MRI), is critical for the diagnosis and stratification of patients to appropriate treatments. Radiomic analyses have the potential to improve the standard of care for patients with BM by applying artificial intelligence (AI) with already acquired medical images to predict clinical outcomes and direct the personalized care of BM patients. Herein, we outline the existing literature applying radiomics for the clinical management of BM. This includes predicting patient response to radiotherapy and identifying radiation necrosis, performing virtual biopsies to predict tumor mutation status, and determining the cancer of origin in brain tumors identified via imaging. With further development, radiomics has the potential to aid in BM patient stratification while circumventing the need for invasive tissue sampling, particularly for patients not eligible for surgical resection.

Importance of sex and gender factors for COVID-19 infection and hospitalisation: a sex-stratified analysis using machine learning in UK Biobank data.

OBJECTIVE: To examine sex and gender roles in COVID-19 test positivity and hospitalisation in sex-stratified predictive models using machine learning. DESIGN: Cross-sectional study. SETTING: UK Biobank prospective cohort. PARTICIPANTS: Participants tested between 16 March 2020 and 18 May 2020 were analysed. MAIN OUTCOME MEASURES: The endpoints of the study were COVID-19 test positivity and hospitalisation. Forty-two individuals' demographics, psychosocial factors and comorbidities were used as likely determinants of outcomes. Gradient boosting machine was used for building prediction models. RESULTS: Of 4510 individuals tested (51.2% female, mean age=68.5±8.9 years), 29.4% tested positive. Males were more likely to be positive than females (31.6% vs 27.3%, p=0.001). In females, living in more deprived areas, lower income, increased low-density lipoprotein (LDL) to high-density lipoprotein (HDL) ratio, working night shifts and living with a greater number of family members were associated with a higher likelihood of COVID-19 positive test. While in males, greater body mass index and LDL to HDL ratio were the factors associated with a positive test. Older age and adverse cardiometabolic characteristics were the most prominent variables associated with hospitalisation of test-positive patients in both overall and sex-stratified models. CONCLUSION: High-risk jobs, crowded living arrangements and living in deprived areas were associated with increased COVID-19 infection in females, while high-risk cardiometabolic characteristics were more influential in males. Gender-related factors have a greater impact on females; hence, they should be considered in identifying priority groups for COVID-19 infection vaccination campaigns.

Radiomics and machine learning for the diagnosis of pediatric cervical non-tuberculous mycobacterial lymphadenitis.

Non-tuberculous mycobacterial (NTM) infection is an emerging infectious entity that often presents as lymphadenitis in the pediatric age group. Current practice involves invasive testing and excisional biopsy to diagnose NTM lymphadenitis. In this study, we performed a retrospective analysis of 249 lymph nodes selected from 143 CT scans of pediatric patients presenting with lymphadenopathy at the Montreal Children's Hospital between 2005 and 2018. A Random Forest classifier was trained on the ten most discriminative features from a set of 1231 radiomic features. The model classifying nodes as pyogenic, NTM, reactive, or proliferative lymphadenopathy achieved an accuracy of 72%, a precision of 68%, and a recall of 70%. Between NTM and all other causes of lymphadenopathy, the model achieved an area under the curve (AUC) of 89%. Between NTM and pyogenic lymphadenitis, the model achieved an AUC of 90%. Between NTM and the reactive and proliferative lymphadenopathy groups, the model achieved an AUC of 93%. These results indicate that radiomics can achieve a high accuracy for classification of NTM lymphadenitis. Such a non-invasive highly accurate diagnostic approach has the potential to reduce the need for invasive procedures in the pediatric population.

Above and Beyond Age: Prediction of Major Postoperative Adverse Events in Head and Neck Surgery.

OBJECTIVE: Major postoperative adverse events (MPAEs) following head and neck surgery are not infrequent and lead to significant morbidity. The objective of this study was to ascertain which factors are most predictive of MPAEs in patients undergoing head and neck surgery. METHODS: A cohort study was carried out based on data from patients registered in the National Surgical Quality Improvement Program (NSQIP) from 2006 to 2018. All patients undergoing non-ambulatory head and neck surgery based on Current Procedural Terminology codes were included. Perioperative factors were evaluated to predict MPAEs within 30-days of surgery. Age was classified as both a continuous and categorical variable. Retained factors were classified by attributable fraction and C-statistic. Multivariate regression and supervised machine learning models were used to quantify the contribution of age as a predictor of MPAEs. RESULTS: A total of 43 701 operations were analyzed with 5106 (11.7%) MPAEs. The results of supervised machine learning indicated that prolonged surgeries, anemia, free tissue transfer, weight loss, wound classification, hypoalbuminemia, wound infection, tracheotomy (concurrent with index head and neck surgery), American Society of Anesthesia (ASA) class, and sex as most predictive of MPAEs. On multivariate regression, ASA class (21.3%), hypertension on medication (15.8%), prolonged operative time (15.3%), sex (13.1%), preoperative anemia (12.8%), and free tissue transfer (9%) had the largest attributable fractions associated with MPAEs. Age was independently associated with MPAEs with an attributable fraction ranging from 0.6% to 4.3% with poor predictive ability (C-statistic 0.60). CONCLUSION: Surgical, comorbid, and frailty-related factors were most predictive of short-term MPAEs following head and neck surgery. Age alone contributed a small attributable fraction and poor prediction of MPAEs. LEVEL OF EVIDENCE: 3.

Silver: Forging almost Gold Standard Datasets.

Gene set analysis has been widely used to gain insight from high-throughput expression studies. Although various tools and methods have been developed for gene set analysis, there is no consensus among researchers regarding best practice(s). Most often, evaluation studies have reported contradictory recommendations of which methods are superior. Therefore, an unbiased quantitative framework for evaluations of gene set analysis methods will be valuable. Such a framework requires gene expression datasets where enrichment status of gene sets is known a priori. In the absence of such gold standard datasets, artificial datasets are commonly used for evaluations of gene set analysis methods; however, they often rely on oversimplifying assumptions that make them biased in favor of or against a given method. In this paper, we propose a quantitative framework for evaluation of gene set analysis methods by synthesizing expression datasets using real data, without relying on oversimplifying or unrealistic assumptions, while preserving complex gene-gene correlations and retaining the distribution of expression values. The utility of the quantitative approach is shown by evaluating ten widely used gene set analysis methods. An implementation of the proposed method is publicly available. We suggest using Silver to evaluate existing and new gene set analysis methods. Evaluation using Silver provides a better understanding of current methods and can aid in the development of gene set analysis methods to achieve higher specificity without sacrificing sensitivity.

Site-Specific Variation in Radiomic Features of Head and Neck Squamous Cell Carcinoma and Its Impact on Machine Learning Models.

Current radiomic studies of head and neck squamous cell carcinomas (HNSCC) are typically based on datasets combining tumors from different locations, assuming that the radiomic features are similar based on histopathologic characteristics. However, molecular pathogenesis and treatment in HNSCC substantially vary across different tumor sites. It is not known if a statistical difference exists between radiomic features from different tumor sites and how they affect machine learning model performance in endpoint prediction. To answer these questions, we extracted radiomic features from contrast-enhanced neck computed tomography scans (CTs) of 605 patients with HNSCC originating from the oral cavity, oropharynx, and hypopharynx/larynx. The difference in radiomic features of tumors from these sites was assessed using statistical analyses and Random Forest classifiers on the radiomic features with 10-fold cross-validation to predict tumor sites, nodal metastasis, and HPV status. We found statistically significant differences (p-value ≤ 0.05) between the radiomic features of HNSCC depending on tumor location. We also observed that differences in quantitative features among HNSCC from different locations impact the performance of machine learning models. This suggests that radiomic features may reveal biologic heterogeneity complementary to current gold standard histopathologic evaluation. We recommend considering tumor site in radiomic studies of HNSCC.

Higher concentrations of vitamin D in Canadian children with juvenile idiopathic arthritis compared to healthy controls are associated with more frequent use of vitamin D supplements and season of birth.

A number of studies have demonstrated that patients with autoimmune disease have lower levels of vitamin D prompting speculation that vitamin D might suppress inflammation and immune responses in children with juvenile idiopathic arthritis (JIA).  The objective of this study was to compare vitamin D levels in children with JIA at disease onset with healthy children. We hypothesized that children and adolescents with JIA have lower vitamin D levels than healthy children and adolescents. Data from a Canadian cohort of children with new-onset JIA (n= 164, data collection 2007-2012) were compared to Canadian Health Measures Survey (CHMS) data (n=4027, data collection 2007-2011). We compared 25-hydroxy vitamin D (25(OH)D) concentrations with measures of inflammation, vitamin D supplement use, milk intake, and season of birth. Mean 25(OH)D level was significantly higher in patients with JIA (79 ± 3.1 nmol/L) than in healthy controls (68 ± 1.8 nmol/L P <.05). Patients with JIA more often used vitamin D containing supplements (50% vs. 7%; P <.05). The prevalence of 25(OH)D deficiency (<30 nmol/L) was 6% for both groups. Children with JIA with 25(OH)D deficiency or insufficiency (<50 nmol/L) had higher C-reactive protein levels. Children with JIA were more often born in the fall and winter compared to healthy children. In contrast to earlier studies, we found vitamin D levels in Canadian children with JIA were higher compared to healthy children and associated with more frequent use of vitamin D supplements. Among children with JIA, low vitamin D levels were associated with indicators of greater inflammation.

Juxtapose: a gene-embedding approach for comparing co-expression networks.

BACKGROUND: Gene co-expression networks (GCNs) are not easily comparable due to their complex structure. In this paper, we propose a tool, Juxtapose, together with similarity measures that can be utilized for comparative transcriptomics between a set of organisms. While we focus on its application to comparing co-expression networks across species in evolutionary studies, Juxtapose is also generalizable to co-expression network comparisons across tissues or conditions within the same species. METHODS: A word embedding strategy commonly used in natural language processing was utilized in order to generate gene embeddings based on walks made throughout the GCNs. Juxtapose was evaluated based on its ability to embed the nodes of synthetic structures in the networks consistently while also generating biologically informative results. Evaluation of the techniques proposed in this research utilized RNA-seq datasets from GTEx, a multi-species experiment of prefrontal cortex samples from the Gene Expression Omnibus, as well as synthesized datasets. Biological evaluation was performed using gene set enrichment analysis and known gene relationships in literature. RESULTS: We show that Juxtapose is capable of globally aligning synthesized networks as well as identifying areas that are conserved in real gene co-expression networks without reliance on external biological information. Furthermore, output from a matching algorithm that uses cosine distance between GCN embeddings is shown to be an informative measure of similarity that reflects the amount of topological similarity between networks. CONCLUSIONS: Juxtapose can be used to align GCNs without relying on known biological similarities and enables post-hoc analyses using biological parameters, such as orthology of genes, or conserved or variable pathways. AVAILABILITY: A development version of the software used in this paper is available at https://github.com/klovens/juxtapose.

Brief History of Artificial Intelligence.

This article reviews the history of artificial intelligence and introduces the reader to major events that prompted interest in the field, as well as pitfalls and challenges that have slowed its development. The purpose of this article is to provide a high-level historical perspective on the development of the field over the past decades, highlighting the potential of the field for transforming health care, but also the importance of setting realistic expectations for artificial intelligence applications to avoid repeating historical cyclical trends and a third "artificial intelligence winter."

Overview of Machine Learning: Part 2: Deep Learning for Medical Image Analysis.

Deep learning has contributed to solving complex problems in science and engineering. This article provides the fundamental background required to understand and develop deep learning models for medical imaging applications. The authors review the main deep learning architectures such as multilayer perceptron, convolutional neural networks, autoencoders, recurrent neural networks, and generative adversarial neural networks. They also discuss the strategies for training deep learning models when the available datasets are imbalanced or of limited size and conclude with a discussion of the obstacles and challenges hindering the deployment of deep learning solutions in clinical settings.

Machine Learning Algorithm Validation: From Essentials to Advanced Applications and Implications for Regulatory Certification and Deployment.

The deployment of machine learning (ML) models in the health care domain can increase the speed and accuracy of diagnosis and improve treatment planning and patient care. Translating academic research to applications that are deployable in clinical settings requires the ability to generalize and high reproducibility, which are contingent on a rigorous and sound methodology for the development and evaluation of ML models. This article describes the fundamental concepts and processes for ML model evaluation and highlights common workflows. It concludes with a discussion of the requirements for the deployment of ML models in clinical settings.

Machine Learning Applications for Head and Neck Imaging.

The head and neck (HN) consists of a large number of vital anatomic structures within a compact area. Imaging plays a central role in the diagnosis and management of major disorders affecting the HN. This article reviews the recent applications of machine learning (ML) in HN imaging with a focus on deep learning approaches. It categorizes ML applications in HN imaging into deep learning and traditional ML applications and provides examples of each category. It also discusses the main challenges facing the successful deployment of ML-based applications in the clinical setting and provides suggestions for addressing these challenges.

Overview of Machine Learning Part 1: Fundamentals and Classic Approaches.

The extensive body of research and advances in machine learning (ML) and the availability of a large volume of patient data make ML a powerful tool for producing models with the potential for widespread deployment in clinical settings. This article provides an overview of the classic supervised and unsupervised ML methods as well as fundamental concepts required for understanding how to develop generalizable and high-performing ML applications. It also describes the important steps for developing a ML model and how decisions made in these steps affect model performance and ability to generalize.

Dual Energy Computed Tomography in Head and Neck Imaging: Pushing the Envelope.

Multiple applications of dual energy computed tomography (DECT) have been described for the evaluation of disorders in the head and neck, especially in oncology. We review the body of evidence suggesting advantages of DECT for the evaluation of the neck compared with conventional single energy computed tomography scans, but the full potential of DECT is still to be realized. There is early evidence suggesting significant advantages of DECT for the extraction of quantitative biomarkers using radiomics and machine learning, representing a new horizon that may enable this technology to reach its full potential.