Improving automatic delineation for head and neck organs at risk by Deep Learning Contouring.

INTRODUCTION: Adequate head and neck (HN) organ-at-risk (OAR) delineation is crucial for HN radiotherapy and for investigating the relationships between radiation dose to OARs and radiation-induced side effects. The automatic contouring algorithms that are currently in clinical use, such as atlas-based contouring (ABAS), leave room for improvement. The aim of this study was to use a comprehensive evaluation methodology to investigate the performance of HN OAR auto-contouring when using deep learning contouring (DLC), compared to ABAS. METHODS: The DLC neural network was trained on 589 HN cancer patients. DLC was compared to ABAS by providing each method with an independent validation cohort of 104 patients, which had also been manually contoured. For each of the 22 OAR contours - glandular, upper digestive tract and central nervous system (CNS)-related structures - the dice similarity coefficient (DICE), and absolute mean and max dose differences (|Δmean-dose| and |Δmax-dose|) performance measures were obtained. For a subset of 7 OARs, an evaluation of contouring time, inter-observer variation and subjective judgement was performed. RESULTS: DLC resulted in equal or significantly improved quantitative performance measures in 19 out of 22 OARs, compared to the ABAS (DICE/|Δmean dose|/|Δmax dose|: 0.59/4.2/4.1 Gy (ABAS); 0.74/1.1/0.8 Gy (DLC)). The improvements were mainly for the glandular and upper digestive tract OARs. DLC significantly reduced the delineation time for the inexperienced observer. The subjective evaluation showed that DLC contours were more often preferable to the ABAS contours overall, were considered to be more precise, and more often confused with manual contours. Manual contours still outperformed both DLC and ABAS; however, DLC results were within or bordering the inter-observer variability for the manual edited contours in this cohort. CONCLUSION: The DLC, trained on a large HN cancer patient cohort, outperformed the ABAS for the majority of HN OARs.

An Evaluation of Atlas Selection Methods for Atlas-Based Automatic Segmentation in Radiotherapy Treatment Planning.

Atlas-based automatic segmentation is used in radiotherapy planning to accelerate the delineation of organs at risk (OARs). Atlas selection has been proposed as a way to improve the accuracy and execution time of segmentation, assuming that, the more similar the atlas is to the patient, the better the results will be. This paper presents an analysis of atlas selection methods in the context of radiotherapy treatment planning. For a range of commonly contoured OARs, a thorough comparison of a large class of typical atlas selection methods has been performed. For this evaluation, clinically contoured CT images of the head and neck ( N=316 ) and thorax ( N=280 ) were used. The state-of-the-art intensity and deformation similarity-based atlas selection methods were found to compare poorly to perfect atlas selection. Counter-intuitively, atlas selection methods based on a fixed set of representative atlases outperformed atlas selection methods based on the patient image. This study suggests that atlas-based segmentation with currently available selection methods compares poorly to the potential best performance, hampering the clinical utility of atlas-based segmentation. Effective atlas selection remains an open challenge in atlas-based segmentation for radiotherapy planning.

Can Atlas-Based Auto-Segmentation Ever Be Perfect? Insights From Extreme Value Theory.

Atlas-based segmentation is used in radiotherapy planning to accelerate the delineation of organs at risk (OARs). Atlas selection has been proposed to improve the performance of segmentation, assuming that the more similar the atlas is to the patient, the better the result. It follows that the larger the database of atlases from which to select, the better the results should be. This paper seeks to estimate a clinically achievable expected performance under this assumption. Assuming a perfect atlas selection, an extreme value theory has been applied to estimate the accuracy of single-atlas and multi-atlas segmentation given a large database of atlases. For this purpose, clinical contours of most common OARs on computed tomography of the head and neck ( N=316 ) and thoracic ( N=280 ) cases were used. This paper found that while for most organs, perfect segmentation cannot be reasonably expected, auto-contouring performance of a level corresponding to clinical quality could be consistently expected given a database of 5000 atlases under the assumption of perfect atlas selection.

Comparative evaluation of autocontouring in clinical practice: A practical method using the Turing test.

PURPOSE: Automated techniques for estimating the contours of organs and structures in medical images have become more widespread and a variety of measures are available for assessing their quality. Quantitative measures of geometric agreement, for example, overlap with a gold-standard delineation, are popular but may not predict the level of clinical acceptance for the contouring method. Therefore, surrogate measures that relate more directly to the clinical judgment of contours, and to the way they are used in routine workflows, need to be developed. The purpose of this study is to propose a method (inspired by the Turing Test) for providing contour quality measures that directly draw upon practitioners' assessments of manual and automatic contours. This approach assumes that an inability to distinguish automatically produced contours from those of clinical experts would indicate that the contours are of sufficient quality for clinical use. In turn, it is anticipated that such contours would receive less manual editing prior to being accepted for clinical use. In this study, an initial assessment of this approach is performed with radiation oncologists and therapists. METHODS: Eight clinical observers were presented with thoracic organ-at-risk contours through a web interface and were asked to determine if they were automatically generated or manually delineated. The accuracy of the visual determination was assessed, and the proportion of contours for which the source was misclassified recorded. Contours of six different organs in a clinical workflow were for 20 patient cases. The time required to edit autocontours to a clinically acceptable standard was also measured, as a gold standard of clinical utility. Established quantitative measures of autocontouring performance, such as Dice similarity coefficient with respect to the original clinical contour and the misclassification rate accessed with the proposed framework, were evaluated as surrogates of the editing time measured. RESULTS: The misclassification rates for each organ were: esophagus 30.0%, heart 22.9%, left lung 51.2%, right lung 58.5%, mediastinum envelope 43.9%, and spinal cord 46.8%. The time savings resulting from editing the autocontours compared to the standard clinical workflow were 12%, 25%, 43%, 77%, 46%, and 50%, respectively, for these organs. The median Dice similarity coefficients between the clinical contours and the autocontours were 0.46, 0.90, 0.98, 0.98, 0.94, and 0.86, respectively, for these organs. CONCLUSIONS: A better correspondence with time saving was observed for the misclassification rate than the quantitative contour measures explored. From this, we conclude that the inability to accurately judge the source of a contour indicates a reduced need for editing and therefore a greater time saving overall. Hence, task-based assessments of contouring performance may be considered as an additional way of evaluating the clinical utility of autosegmentation methods.

Myocardial strain computed at multiple spatial scales from tagged magnetic resonance imaging: Estimating cardiac biomarkers for CRT patients.

Abnormal cardiac motion can indicate different forms of disease, which can manifest at different spatial scales in the myocardium. Many studies have sought to characterise particular motion abnormalities associated with specific diseases, and to utilise motion information to improve diagnoses. However, the importance of spatial scale in the analysis of cardiac deformation has not been extensively investigated. We build on recent work on the analysis of myocardial strains at different spatial scales using a cardiac motion atlas to find the optimal scales for estimating different cardiac biomarkers. We apply a multi-scale strain analysis to a 43 patient cohort of cardiac resynchronisation therapy (CRT) patients using tagged magnetic resonance imaging data for (1) predicting response to CRT, (2) identifying septal flash, (3) estimating QRS duration, and (4) identifying the presence of ischaemia. A repeated, stratified cross-validation is used to demonstrate the importance of spatial scale in our analysis, revealing different optimal spatial scales for the estimation of different biomarkers.

Clinical evaluation of atlas and deep learning based automatic contouring for lung cancer.

BACKGROUND AND PURPOSE: Contouring of organs at risk (OARs) is an important but time consuming part of radiotherapy treatment planning. The aim of this study was to investigate whether using institutional created software-generated contouring will save time if used as a starting point for manual OAR contouring for lung cancer patients. MATERIAL AND METHODS: Twenty CT scans of stage I-III NSCLC patients were used to compare user adjusted contours after an atlas-based and deep learning contour, against manual delineation. The lungs, esophagus, spinal cord, heart and mediastinum were contoured for this study. The time to perform the manual tasks was recorded. RESULTS: With a median time of 20 min for manual contouring, the total median time saved was 7.8 min when using atlas-based contouring and 10 min for deep learning contouring. Both atlas based and deep learning adjustment times were significantly lower than manual contouring time for all OARs except for the left lung and esophagus of the atlas based contouring. CONCLUSIONS: User adjustment of software generated contours is a viable strategy to reduce contouring time of OARs for lung radiotherapy while conforming to local clinical standards. In addition, deep learning contouring shows promising results compared to existing solutions.

Cardiac and Respiratory Motion Correction for Simultaneous Cardiac PET/MR.

Cardiac PET is a versatile imaging technique providing important diagnostic information about ischemic heart diseases. Respiratory and cardiac motion of the heart can strongly impair image quality and therefore diagnostic accuracy of cardiac PET scans. The aim of this study was to investigate a new cardiac PET/MR approach providing respiratory and cardiac motion-compensated MR and PET images in less than 5 min. Methods: Free-breathing 3-dimensional MR data were acquired and retrospectively binned into multiple respiratory and cardiac motion states. Three-dimensional cardiac and respiratory motion fields were obtained with a nonrigid registration algorithm and used in motion-compensated MR and PET reconstructions to improve image quality. The improvement in image quality and diagnostic accuracy of the technique was assessed in simultaneous 18F-FDG PET/MR scans of a canine model of myocardial infarct and was demonstrated in a human subject. Results: MR motion fields were successfully used to compensate for in vivo cardiac motion, leading to improvements in full width at half maximum of the canine myocardium of 13% ± 5%, similar to cardiac gating but with a 90% ± 57% higher contrast-to-noise ratio between myocardium and blood. Motion correction led to an improvement in MR image quality in all subjects, with an increase in sharpness of the canine coronary arteries of 85% ± 72%. A functional assessment showed good agreement with standard MR cine scans with a difference in ejection fraction of -2% ± 3%. MR-based respiratory and cardiac motion information was used to improve the PET image quality of a human in vivo scan. Conclusion: The MR technique presented here provides both diagnostic and motion information that can be used to improve MR and PET image quality. Reliable respiratory and cardiac motion correction could make cardiac PET results more reproducible.

Registration of Multiview Echocardiography Sequences Using a Subspace Error Metric.

OBJECTIVE: 3-D +t echocardiography (3DtE) is widely employed for the assessment of left ventricular anatomy and function. However, the information derived from 3DtE images can be affected by the poor image quality and the limited field of view. Registration of multiview 3DtE sequences has been proposed to compound images from different acoustic windows, therefore improving both image quality and coverage. We propose a novel subspace error metric for an automatic and robust registration of multiview intrasubject 3DtE sequences. METHODS: The proposed metric employs linear dimensionality reduction to exploit the similarity in the temporal variation of multiview 3DtE sequences. The use of a low-dimensional subspace for the computation of the error metric reduces the influence of image artefacts and noise on the registration optimization, resulting in fast and robust registrations that do not require a starting estimate. RESULTS: The accuracy, robustness, and execution time of the proposed registration were thoroughly validated. Results on 48 pairwise multiview 3DtE registrations show the proposed error metric to outperform a state-of-the-art phase-based error metric, with improvements in median/75th percentile of the target registration error of 21%/31% and an improvement in mean execution time of 45%. CONCLUSION: The proposed subspace error metric outperforms sum-of-squared differences and phase-based error metrics for the registration of multiview 3DtE sequences in terms of accuracy, robustness, and execution time. SIGNIFICANCE: The use of the proposed subspace error metric has the potential to replace standard image error metrics for a robust and automatic registration of multiview 3DtE sequences.

Non-invasive Model-Based Assessment of Passive Left-Ventricular Myocardial Stiffness in Healthy Subjects and in Patients with Non-ischemic Dilated Cardiomyopathy.

Patient-specific modelling has emerged as a tool for studying heart function, demonstrating the potential to provide non-invasive estimates of tissue passive stiffness. However, reliable use of model-derived stiffness requires sufficient model accuracy and unique estimation of model parameters. In this paper we present personalised models of cardiac mechanics, focusing on improving model accuracy, while ensuring unique parametrisation. The influence of principal model uncertainties on accuracy and parameter identifiability was systematically assessed in a group of patients with dilated cardiomyopathy ([Formula: see text]) and healthy volunteers ([Formula: see text]). For all cases, we examined three circumferentially symmetric fibre distributions and two epicardial boundary conditions. Our results demonstrated the ability of data-derived boundary conditions to improve model accuracy and highlighted the influence of the assumed fibre distribution on both model fidelity and stiffness estimates. The model personalisation pipeline-based strictly on non-invasive data-produced unique parameter estimates and satisfactory model errors for all cases, supporting the selected model assumptions. The thorough analysis performed enabled the comparison of passive parameters between volunteers and dilated cardiomyopathy patients, illustrating elevated stiffness in diseased hearts.

A framework for combining a motion atlas with non-motion information to learn clinically useful biomarkers: Application to cardiac resynchronisation therapy response prediction.

We present a framework for combining a cardiac motion atlas with non-motion data. The atlas represents cardiac cycle motion across a number of subjects in a common space based on rich motion descriptors capturing 3D displacement, velocity, strain and strain rate. The non-motion data are derived from a variety of sources such as imaging, electrocardiogram (ECG) and clinical reports. Once in the atlas space, we apply a novel supervised learning approach based on random projections and ensemble learning to learn the relationship between the atlas data and some desired clinical output. We apply our framework to the problem of predicting response to Cardiac Resynchronisation Therapy (CRT). Using a cohort of 34 patients selected for CRT using conventional criteria, results show that the combination of motion and non-motion data enables CRT response to be predicted with 91.2% accuracy (100% sensitivity and 62.5% specificity), which compares favourably with the current state-of-the-art in CRT response prediction.

Personalising population-based respiratory motion models of the heart using neighbourhood approximation based on learnt anatomical features.

Respiratory motion models have been proposed for the estimation and compensation of respiratory motion during image acquisition and image-guided interventions on organs in the chest and abdomen. However, such techniques are not commonly used in the clinic. Subject-specific motion models require a dynamic calibration scan that interrupts the clinical workflow and is often impractical to acquire, while population-based motion models are not as accurate as subject-specific motion models. To address this lack of accuracy, we propose a novel personalisation framework for population-based respiratory motion models and demonstrate its application to respiratory motion of the heart. The proposed method selects a subset of the population sample which is more likely to represent the cardiac respiratory motion of an unseen subject, thus providing a more accurate motion model. The selection is based only on anatomical features of the heart extracted from a static image. The features used are learnt using a neighbourhood approximation technique from a set of training datasets for which respiratory motion estimates are available. Results on a population sample of 28 adult healthy volunteers show average improvements in estimation accuracy of 20% compared to a standard population-based motion model, with an average value for the 50th and 95th quantiles of the estimation error of 1.6mm and 4.7 mm respectively. Furthermore, the anatomical features of the heart most strongly correlated to respiratory motion are investigated for the first time, showing the features on the apex in proximity to the diaphragm and the rib cage, on the left ventricle and interventricular septum to be good predictors of the similarity in cardiac respiratory motion.

A novel Bayesian respiratory motion model to estimate and resolve uncertainty in image-guided cardiac interventions.

In image-guided cardiac interventions, respiratory motion causes misalignments between the pre-procedure roadmap of the heart used for guidance and the intra-procedure position of the heart, reducing the accuracy of the guidance information and leading to potentially dangerous consequences. We propose a novel technique for motion-correcting the pre-procedural information that combines a probabilistic MRI-derived affine motion model with intra-procedure real-time 3D echocardiography (echo) images in a Bayesian framework. The probabilistic model incorporates a measure of confidence in its motion estimates which enables resolution of the potentially conflicting information supplied by the model and the echo data. Unlike models proposed so far, our method allows the final motion estimate to deviate from the model-produced estimate according to the information provided by the echo images, so adapting to the complex variability of respiratory motion. The proposed method is evaluated using gold-standard MRI-derived motion fields and simulated 3D echo data for nine volunteers and real 3D live echo images for four volunteers. The Bayesian method is compared to 5 other motion estimation techniques and results show mean/max improvements in estimation accuracy of 10.6%/18.9% for simulated echo images and 20.8%/41.5% for real 3D live echo data, over the best comparative estimation method.