Search

This editorial introduces the Special Issue on Simulation and Synthesis in Medical Imaging. In this editorial, we define so-far ambiguous terms of simulation and synthesis in medical imaging. We also briefly discuss the synergistic importance of mechanistic (hypothesis-driven) and phenomenological (data-driven) models of medical image generation. Finally, we introduce the twelve papers published in this issue covering both mechanistic (5) and phenomenological (7) medical image generation. This rich selection of papers covers applications in cardiology, retinopathy, histopathology, neurosciences, and oncology. It also covers all mainstream diagnostic medical imaging modalities. We conclude the editorial with a personal view on the field and highlight some existing challenges and future research opportunities. [ABSTRACT FROM PUBLISHER]

Medical image analysis has grown into a matured field challenged by progress made across all medical imaging technologies and more recent breakthroughs in biological imaging. The cross-fertilisation between medical image analysis, biomedical imaging physics and technology, and domain knowledge from medicine and biology has spurred a truly interdisciplinary effort that stretched outside the original boundaries of the disciplines that gave birth to this field and created stimulating and enriching synergies. Consideration on how the field has evolved and the experience of the work carried out over the last 15 years in our centre, has led us to envision a future emphasis of medical imaging in Precision Imaging. Precision Imaging is not a new discipline but rather a distinct emphasis in medical imaging borne at the cross-roads between, and unifying the efforts behind mechanistic and phenomenological model-based imaging. It captures three main directions in the effort to deal with the information deluge in imaging sciences, and thus achieve wisdom from data, information, and knowledge. Precision Imaging is finally characterised by being descriptive, predictive and integrative about the imaged object. This paper provides a brief and personal perspective on how the field has evolved, summarises and formalises our vision of Precision Imaging for Precision Medicine, and highlights some connections with past research and current trends in the field. [ABSTRACT FROM AUTHOR]

An introduction is presented in which the editor discusses various papers within the issue on topics including biomedical engineering, computational biology and the role of ontologies in biology.

Integrative models of cardiac physiology are important for understanding disease and planning intervention. Multimodal cardiovascular imaging plays an important role in defining the computational domain, the boundary/initial conditions, and tissue function and properties. Computational models can then be personalized through information derived from in vivo and, when possible, non-invasive images. Efforts are now established to provide Web-accessible structural and functional atlases of the normal and pathological heart for clinical, research and educational purposes. Efficient and robust statistical representations of cardiac morphology and morphodynamics can thereby be obtained, enabling quantitative analysis of images based on such representations. Statistical models of shape and appearance can be built automatically from large populations of image datasets by minimizing manual intervention and data collection. These methods facilitate statistical analysis of regional heart shape and wall motion characteristics across population groups, via the application of parametric mathematical modelling tools. These parametric modelling tools and associated ontological schema also facilitate data fusion between different imaging protocols and modalities as well as other data sources. Statistical priors can also be used to support cardiac image analysis with applications to advanced quantification and subject-specific simulations of computational physiology. [ABSTRACT FROM AUTHOR]

Abstract-In this paper, a statistical shape analysis method for myocardial contraction is presented that was built to detect and locate regional wall motion abnormalities (RWMA). For each slice level (base, middle, and apex), 44 short-axis magnetic resonance images were selected from healthy volunteers to train a statistical model of normal myocardial contraction using independent component analysis (ICA). A classification algorithm was constructed from the ICA components to automatically detect and localize abnormally contracting regions of the myocardium. The algorithm was validated on 45 patients suffering from ischemic heart disease. Two validations were performed; one with visual wall motion scores (VWMS) and the other with wall thickening (WT) used as references. Accuracy of the ICA-based method on each slice level was 69.93% (base), 89.63% (middle), and 71.78% (apex) when WT was used as reference, and 63.70% (base), 67.41% (middle), and 66.67% (apex) when VWMS was used as reference. From this we conclude that the proposed method is a promising diagnostic support tool to assist clinicians in reducing the subjectivity in VWMS. [ABSTRACT FROM AUTHOR]

One of the drawbacks of statistical shape models is their occasional failure to converge. Although visually this fact is usually easy to recognize, there is no automatic way to detect it. In this paper, we introduce a generic reliability measure for statistical shape models. It is based on a probabilistic framework and uses information extracted by the model itself during the matching process. The proposed method was validated with two variants of Active Shape Models in the context facial image analysis. Experimental results on more than 3700 facial images showed a high degree of correlation between the segmentation accuracy and the estimated reliability metric. [ABSTRACT FROM AUTHOR]

Abstract: Background: Modifying lipids levels underpins atherosclerosis prevention. Flow-mediated dilation (FMD) could advise which patients to treat and to what extent. Little is known about the influence of near-normal lipid levels on the endothelium and the mechanisms related to different lipid fractions. We studied associations between FMD and lipids, focusing on normal lipid levels. Methods: An age-homogeneous sample of 171 healthy, untreated military men (mean age 35.5±1.1 years) was studied: serum lipid determination and brachial artery ultrasound with a forearm ischemia cuff and automated measurement were performed. NCEP-ATP III groups were used. Results: Significantly smaller vessel diameters were found among individuals with high HDL-c (4.10 mm vs. 4.24 mm), optimal LDL-c (4.00 mm vs. 4.22 mm), and normal triglycerides (<150 mg/dl) (4.15 mm vs. 4.31 mm). Basal diameter correlated significantly with HDL-c and triglycerides. There were significant differences in FMD between low HDL-c compared to the rest (4.13% vs. 5.07%) and between optimal and near-optimal LDL-c compared to the rest (5.28% vs. 4.56%). HDL-c and LDL-c correlated with FMD. The inverse relation of high LDL-c and FMD is partially due to a decreased stimulus. Besides, stimulus heterogeneity may mask HDL-c link with FMD. Conclusion: Those subjects naturally (not pharmacologically) in the healthy tail-end of the lipid distributions have the best endothelial function and smaller vessels. Functional vascular remodeling might precede anatomical remodeling and, in early stages, vessel size should be considered a risk indicator rather than an atherosclerotic sign. Furthermore, controlling the stimulus seems necessary for detecting the relationship between HDL-c and FMD, and should be performed regularly. [Copyright &y& Elsevier]

Abstract: Segmentation of vascular structures is a difficult and challenging task. In this article, we present an algorithm devised for the segmentation of such structures. Our technique consists in a geometric deformable model with associated energy functional that incorporates high-order multiscale features in a non-parametric statistical framework. Although the proposed segmentation method is generic, it has been applied to the segmentation of cerebral aneurysms in 3DRA and CTA. An evaluation study over 10 clinical datasets indicate that the segmentations obtained by our method present a high overlap index with respect to the ground-truth (91.13% and 73.31%, respectively) and that the mean error distance from the surface to the ground truth is close to the in-plane resolution (0.40 and 0.38mm, respectively). Besides, our technique favorably compares to other alternative techniques based on deformable models, namely parametric geodesic active regions and active contours without edges. [Copyright &y& Elsevier]

This paper presents a framework for weighted fusion of several Active Shape and Active Appearance Models. The approach is based on the eigenspace fusion method proposed by Hall et al. [1], which has been extended to fuse more than two weighted eigenspaces using unbiased mean and covariance matrix estimates. To evaluate the performance of fusion, a comparative assessment on segmentation precision as well as facial verification tests are performed using the AR, EQUINOX, and XM2VTS databases. Based on the results, it is concluded that the fusion is useful when the model needs to be updated online or when the original observations are absent. [ABSTRACT FROM AUTHOR]

This paper examines the theory of kernel Fisher discriminant analysis (KFD) in a Hilbert space and develops a two-phase KFD framework, i.e., kernel principal component analysis (KPCA) plus Fisher linear discriminant analysis (LDA). This framework provides novel insights into the nature of KFD. Based on this framework, the authors propose a complete kernel Fisher discriminant analysis (CKFD) algorithm. CKFD can be used to carry out discriminant analysis in "double discriminant subspaces." The fact that; it can make full use of two kinds of discriminant information, regular arid irregular, makes CKFD a more powerful discriminator. The proposed algorithm was tested and evaluated using the FERET face database and the CENPARMI handwritten numeral database. The experimental results show that CKFD outperforms other KFD algorithms. [ABSTRACT FROM AUTHOR]

Kernel-based methods have been of wide concern in the field of machine learning and neurocomputing. In this paper, a new Kernel Fisher discriminant analysis (KFD) algorithm, called complete KFD (CKFD), is developed. CKFD has two advantages over the existing KFD algorithms. First, its implementation is divided into two phases, i.e., Kernel principal component analysis (KPCA) plus Fisher linear discriminant analysis (FLD), which makes it more transparent and simpler. Second, CKFD can make use of two categories of discriminant information, which makes it more powerful. The proposed algorithm was applied to face recognition and tested on a subset of the FERET database. The experimental results demonstrate that CKFD is significantly better than the algorithms of Kernel Fisherface and Kernel Eigenface. [Copyright &y& Elsevier]

Flow-mediated dilation (FMD) offers a mechanism to characterize endothelial function and, therefore, may play a role in the diagnosis of cardiovascular diseases. Computerized analysis techniques are very desirable to give accuracy and objectivity to the measurements. Virtually all methods proposed up to now to measure FMD rely on accurate edge detection of the arterial wall, and they are not always robust in the presence of poor image quality or image artifacts. A novel method for automatic dilation assessment based on a global image analysis strategy is presented. We model interframe arterial dilation as a superposition of a rigid motion and a scaling factor perpendicular to the artery. Rigid motion can be interpreted as a global compensation for patient and probe movements, an aspect that has not been sufficiently studied before. The scaling factor explains arterial dilation. The ultrasound sequence is analyzed in two phases using image registration to recover both transformation models. Temporal continuity in the registration parameters along the sequence is enforced with a Kalman filter since the dilation process is known to be a gradual physiological phenomenon. Comparing automated and gold standard measurements (average of manual measurements) we found a negligible bias (0.05%FMD) and a small standard deviation (SD) of the differences (1.05%FMD). These values are comparable with those obtained from manual measurements (bias = 0.23%FMD, S[SUBDintra -- obs] = 1.13%FMD, SD[SUBinter -- obs] = 1.20%FMD). The proposed method offers also better reproducibility (CV = 0.40%) than the manual measurements (CV = 1.04%). [ABSTRACT FROM AUTHOR]

In this paper, we show how the concept of statistical deformation models (SDMs) can be used for the construction of average models of the anatomy and their variability. SDMs are built by performing a statistical analysis of the deformations required to map anatomical features in one subject into the corresponding features in another subject. The concept of SDMs is similar to statistical shape models (SSMs) which capture statistical information about shapes across a population, but offers several advantages over SSMs. First, SDMs can be constructed directly from images such as three-dimensional (3-D) magnetic resonance (MR) or computer tomography volumes without the need for segmentation which is usually a prerequisite for the construction of SSMs. Instead, a nonrigid registration algorithm based on free-form deformations and normalized mutual information is used to compute the deformations required to establish dense correspondences between the reference subject and the subjects in the population class under investigation. Second, SDMs allow the construction of an atlas of the average anatomy as well as its variability across a population of subjects. Finally, SDMs take the 3-D nature of the underlying anatomy into account by analysing dense 3-D deformation fields rather than only information about the surface shape of anatomical structures. We show results for the construction of anatomical models of the brain from the MR images of 25 different subjects. The correspondences obtained by the nonrigid registration are evaluated using anatomical landmark locations and show an average error of 1.40 mm at these anatomical landmark positions. We also demonstrate that SDMs can be constructed so as to minimize the bias toward the chosen reference subject. [ABSTRACT FROM AUTHOR]

A novel method is introduced for the generation of landmarks for three-dimensional (3-D) shapes and the construction of the corresponding 3-D statistical shape models. Automatic landmarking of a set of manual segmentations from a class of shapes is achieved by 1) construction of an atlas of the class, 2) automatic extraction of the landmarks from the atlas, and 3) subsequent propagation of these landmarks to each example shape via a volumetric nonrigid registration technique using multiresolution B-spline deformations. This approach presents some advantages over previously published methods: it can treat multiple-part structures and requires less restrictive assumptions on the structure's topology. In this paper, we address the problem of building a 3-D statistical shape model of the left and right ventricle of the heart from 3-D magnetic resonance images. The average accuracy in landmark propagation is shown to be below 2.2 mm. This application demonstrates the robustness and accuracy of the method in the presence of large shape variability and multiple objects. [ABSTRACT FROM AUTHOR]

Focuses on three-dimensional cardiovascular imaging. Diagnostic role; Mortality from cardiovascular diseases, according to the World Health Organization; Congresses on medical imaging.

An active shape model segmentation scheme is presented that is steered by optimal local features, contrary to normalized first order derivative profiles, as in the original formulation [Cootes and Taylor, 1995, 1999, and 2001]. A nonlinear kNN-classifter is used, instead of the linear Mahalanobis distance, to find optimal displacements for landmarks. For each of the landmarks that describe the shape, at each resolution level taken into account during the segmentation optimization procedure, a distinct set of optimal features is determined. The selection of features is automatic, using the training images and sequential feature forward and backward selection. The new approach is tested on synthetic data and in four medical segmentation tasks: segmenting the right and left lung fields in a database of 230 chest radiographs, and segmenting the cerebellum and corpus callosum in a database of 90 slices from MRI brain images. In all cases, the new method produces significantly better results in terms of an overlap error measure (p < 0.001 using a paired T-test) than the original active shape model scheme. [ABSTRACT FROM AUTHOR]

Presents a framework to analyze the propagation of measurement noise through backprojection reconstruction algorithms in electrical impedance tomography (EIT). Sources of measurement noise in EIT acquisition system; Discussion on the influence of the acquisition system architecture; Variants of backprojection reconstruction; Comparison of the results of the error propagation theory.

Focuses on a study which examined three-dimensional modeling for functional analysis of cardiac images. Imaging techniques for cardiac examination; Classical descriptors of cardiac function; Comparison of the performance of different techniques.

Background: Stroke is a leading cause of morbidity and mortality. Retinal imaging allows non-invasive assessment of the microvasculature. Consequently, retinal imaging is a technology which is garnering increasing attention as a means of assessing cardiovascular health and stroke risk. Methods: A biomedical literature search was performed to identify prospective studies that assess the role of retinal imaging derived biomarkers as indicators of stroke risk. Results: Twenty-four studies were included in this systematic review. The available evidence suggests that wider retinal venules, lower fractal dimension, increased arteriolar tortuosity, presence of retinopathy, and presence of retinal emboli are associated with increased likelihood of stroke. There is weaker evidence to suggest that narrower arterioles and the presence of individual retinopathy traits such as microaneurysms and arteriovenous nicking indicate increased stroke risk. Our review identified three models utilizing artificial intelligence algorithms for the analysis of retinal images to predict stroke. Two of these focused on fundus photographs, whilst one also utilized optical coherence tomography (OCT) technology images. The constructed models performed similarly to conventional risk scores but did not significantly exceed their performance. Only two studies identified in this review used OCT imaging, despite the higher dimensionality of this data. Conclusion: Whilst there is strong evidence that retinal imaging features can be used to indicate stroke risk, there is currently no predictive model which significantly outperforms conventional risk scores. To develop clinically useful tools, future research should focus on utilization of deep learning algorithms, validation in external cohorts, and analysis of OCT images. [ABSTRACT FROM AUTHOR]

Proposes a model-based method for quantitative three-dimensional magnetic resonance angiography (MRA). Discussion on the model-based vessel segmentation; In vitro evaluation of the MRA; Illustration of the algorithm on patient data; Analytical model of a toroidal vessel.

• A framework for LRV detection, segmentation in large CMR studies. • Automatic initialization of bi-ventricles for the first time frame without any user intervention. • Automatic End-diastole and End-systole phase assessments. • Supervised deep neural network employed to train and segment the heart. • 3D-ASM image search procedure improved by combining image intensity models with derived distance maps for LV and RV edge localization. The sheer volume of data generated by population imaging studies is unparalleled by current capabilities to extract objective and quantitative cardiac phenotypes; subjective and time-consuming manual image analysis remains the gold standard. Automated image analytics to compute quantitative imaging biomarkers of cardiac function are desperately needed. Data volumes and their variability pose a challenge to most state-of-the-art methods for endo- and epicardial contours, which lack robustness when applied to very large datasets. Our aim is to develop an analysis pipeline for the automatic quantification of cardiac function from cine magnetic resonance imaging data. This work adopt 4,638 cardiac MRI cases coming from UK Biobank with ground truth available for left and RV contours. A hybrid and robust algorithm is proposed to improve the accuracy of automatic left and right ventricle segmentation by harnessing the localization accuracy of deep learning and the morphological accuracy of 3D-ASM (three-dimensional active shape models). The contributions of this paper are three-fold. First, a fully automatic method is proposed for left and right ventricle initialization and cardiac MRI segmentation by taking full advantage of spatiotemporal constraint. Second, a deeply supervised network is introduced to train and segment the heart. Third, the 3D-ASM image search procedure is improved by combining image intensity models with convolutional neural network (CNN) derived distance maps improving endo- and epicardial edge localization. The proposed architecture outperformed the state of the art for cardiac MRI segmentation from UK Biobank. The statistics of RV landmarks detection errors for Triscuspid valve and RV apex are 4.17 mm and 5.58 mm separately. The overlap metric, mean contour distance, Hausdorff distance and cardiac functional parameters are calculated for the LV (Left Ventricle) and RV (Right Ventricle) contour segmentation. Bland–Altman analysis for clinical parameters shows that the results from our automated image analysis pipelines are in good agreement with results from expert manual analysis. Our hybrid scheme combines deep learning and statistical shape modeling for automatic segmentation of the LV/RV from cardiac MRI datasets is effective and robust and can compute cardiac functional indexes from population imaging. [ABSTRACT FROM AUTHOR]

Multi-modality medical imaging is increasingly used for comprehensive assessment of complex diseases in either diagnostic examinations or as part of medical research trials. Different imaging modalities provide complementary information about living tissues. However, multi-modal examinations are not always possible due to adversary factors, such as patient discomfort, increased cost, prolonged scanning time, and scanner unavailability. In additionally, in large imaging studies, incomplete records are not uncommon owing to image artifacts, data corruption or data loss, which compromise the potential of multi-modal acquisitions. In this paper, we propose a weakly coupled and geometry co-regularized joint dictionary learning method to address the problem of cross-modality synthesis while considering the fact that collecting the large amounts of training data is often impractical. Our learning stage requires only a few registered multi-modality image pairs as training data. To employ both paired images and a large set of unpaired data, a cross-modality image matching criterion is proposed. Then, we propose a unified model by integrating such a criterion into the joint dictionary learning and the observed common feature space for associating cross-modality data for the purpose of synthesis. Furthermore, two regularization terms are added to construct robust sparse representations. Our experimental results demonstrate superior performance of the proposed model over state-of-the-art methods. [ABSTRACT FROM PUBLISHER]

Background: Using computational fluid dynamics (CFD) to compute the hemodynamics in cerebral aneurysms has received much attention in the last decade. The usability of these methods depends on the quality of the computations, highlighted in recent discussions. The purpose of this study is to investigate the convergence of common hemodynamic indicators with respect to numerical resolution. Methods: 38 middle cerebral artery bifurcation aneurysms were studied at two different resolutions (one comparable to most studies, and one finer). Relevant hemodynamic indicators were collected from two of the most cited studies, and were compared at the two refinements. In addition, correlation to rupture was investigated. Results: Most of the hemodynamic indicators were very well resolved at the coarser resolutions, correlating with the finest resolution with a correlation coefficient >0.95. The oscillatory shear index (OSI) had the lowest correlation coefficient of 0.83. A logarithmic Bland-Altman plot revealed noticeable variations in the proportion of the aneurysm under low shear, as well as in spatial and temporal gradients not captured by the correlation alone. Conclusion: Statistically, hemodynamic indicators agree well across the different resolutions studied here. However, there are clear outliers visible in several of the hemodynamic indicators, which suggests that special care should be taken when considering individual assessment. [ABSTRACT FROM AUTHOR]

Objectives: Relapse occurs in ~20% of patients with classical Hodgkin lymphoma (cHL) despite treatment adaption based on 2-deoxy-2-[18F]fluoro-D-glucose positron emission tomography/computed tomography response. The objective was to evaluate pre-treatment FDG PET/CT-derived machine learning (ML) models for predicting outcome in patients with cHL.Methods: All cHL patients undergoing pre-treatment PET/CT at our institution between 2008 and 2018 were retrospectively identified. A 1.5 × mean liver standardised uptake value (SUV) and a fixed 4.0 SUV threshold were used to segment PET/CT data. Feature extraction was performed using PyRadiomics with ComBat harmonisation. Training (80%) and test (20%) cohorts stratified around 2-year event-free survival (EFS), age, sex, ethnicity and disease stage were defined. Seven ML models were trained and hyperparameters tuned using stratified 5-fold cross-validation. Area under the curve (AUC) from receiver operator characteristic analysis was used to assess performance.Results: A total of 289 patients (153 males), median age 36 (range 16-88 years), were included. There was no significant difference between training (n = 231) and test cohorts (n = 58) (p value > 0.05). A ridge regression model using a 1.5 × mean liver SUV segmentation had the highest performance, with mean training, validation and test AUCs of 0.82 ± 0.002, 0.79 ± 0.01 and 0.81 ± 0.12. However, there was no significant difference between a logistic model derived from metabolic tumour volume and clinical features or the highest performing radiomic model.Conclusions: Outcome prediction using pre-treatment FDG PET/CT-derived ML models is feasible in cHL patients. Further work is needed to determine optimum predictive thresholds for clinical use.Key Points: • A fixed threshold segmentation method led to more robust radiomic features. • A radiomic-based model for predicting 2-year event-free survival in classical Hodgkin lymphoma patients is feasible. • A predictive model based on ridge regression was the best performing model on our dataset. [ABSTRACT FROM AUTHOR]

Nuclear magnetic resonance (NMR) has proven of enormous value in the investigation of porous media. Its use allows to study pore size distributions, tortuosity, and permeability as a function of the relaxation time, diffusivity, and flow. This information plays an important role in plenty of applications, ranging from oil industry to medical diagnosis. A complete NMR analysis involves the solution of the Bloch–Torrey (BT) equation. However, solving this equation analytically becomes intractable for all but the simplest geometries. We present an efficient numerical framework for solving the complete BT equation in arbitrarily complex domains. In addition to the standard BT equation, the generalised BT formulation takes into account the flow and relaxation terms, allowing a better representation of the phenomena under scope. The presented framework is flexible enough to deal parametrically with any order of convergence in the spatial domain. The major advantage of such approach is to allow both faster computations and sensitivity analyses over realistic geometries. Moreover, we developed a second-order implicit scheme for the temporal discretisation with similar computational demands as the existing explicit methods. This represents a huge step forward for obtaining reliable results with few iterations. Comparisons with analytical solutions and real data show the flexibility and accuracy of the proposed methodology. [ABSTRACT FROM AUTHOR]

• An effective two-stage pipeline is proposed to detect and synthesise absent slices in both the apical and basal region. • The detection model comprises several dense blocks containing ConvLSTM layers, to leverage through-plane contextual and sequential ordering information of slices in cine MR data. • The imputation network is based on a cascaded conditional GAN that can infer multiple missing slices that are anatomically plausible and lead to improved accuracy of subsequent analyses on cardiac MRIs. • The results compensated for the absence of two basal slices show that the mean differences to the reference of stroke volume and ejection fraction are only -1.3 mL and -1.0%, respectively. • The pipeline can improve the reliability of image analysis in large-scale population studies, with no need for re-scanning the patient or discarding incomplete studies. [Display omitted] Cardiac MR acquisition with complete coverage from base to apex is required to ensure accurate subsequent analyses, such as volumetric and functional measurements. However, this requirement cannot be guaranteed when acquiring images in the presence of motion induced by cardiac muscle contraction and respiration. To address this problem, we propose an effective two-stage pipeline for detecting and synthesising absent slices in both the apical and basal region. The detection model comprises several dense blocks containing convolutional long short-term memory (ConvLSTM) layers, to leverage through-plane contextual and sequential ordering information of slices in cine MR data and achieve reliable classification results. The imputation network is based on a dedicated conditional generative adversarial network (GAN) that helps retain key visual cues and fine structural details in the synthesised image slices. The proposed network can infer multiple missing slices that are anatomically plausible and lead to improved accuracy of subsequent analyses on cardiac MRIs, e.g., ventricle segmentation, cardiac quantification compared to those derived from incomplete cardiac MR datasets. For instance, the results obtained when compensating for the absence of two basal slices show that the mean differences to the reference of stroke volume and ejection fraction are only -1.3 mL and -1.0%, respectively, which are significantly smaller than those calculated from the incomplete data (-26.8 mL and -6.7%). The proposed approach can improve the reliability of high-throughput image analysis in large-scale population studies, minimising the need for re-scanning patients or discarding incomplete acquisitions. [ABSTRACT FROM AUTHOR]

Simple Summary: Diffuse large B-cell lymphoma (DLBCL) is the most common type of lymphoma. Even with the improvements in the treatment of DLBCL, around a quarter of patients will experience recurrence. The aim of this single centre retrospective study was to predict which patients would have recurrence within 2 years of their treatment using machine learning techniques based on radiomics extracted from the staging PET/CT images. Our study demonstrated that in our dataset of 229 patients (training data = 183, test data = 46) that a combined radiomic and clinical based model performed better than a simple model based on metabolic tumour volume, and that it had a good predictive ability which was maintained when tested on an unseen test set. Background: Approximately 30% of patients with diffuse large B-cell lymphoma (DLBCL) will have recurrence. The aim of this study was to develop a radiomic based model derived from baseline PET/CT to predict 2-year event free survival (2-EFS). Methods: Patients with DLBCL treated with R-CHOP chemotherapy undergoing pre-treatment PET/CT between January 2008 and January 2018 were included. The dataset was split into training and internal unseen test sets (ratio 80:20). A logistic regression model using metabolic tumour volume (MTV) and six different machine learning classifiers created from clinical and radiomic features derived from the baseline PET/CT were trained and tuned using four-fold cross validation. The model with the highest mean validation receiver operator characteristic (ROC) curve area under the curve (AUC) was tested on the unseen test set. Results: 229 DLBCL patients met the inclusion criteria with 62 (27%) having 2-EFS events. The training cohort had 183 patients with 46 patients in the unseen test cohort. The model with the highest mean validation AUC combined clinical and radiomic features in a ridge regression model with a mean validation AUC of 0.75 ± 0.06 and a test AUC of 0.73. Conclusions: Radiomics based models demonstrate promise in predicting outcomes in DLBCL patients. [ABSTRACT FROM AUTHOR]

Magnetic resonance (MR) images can be corrupted by artifacts and noise, potentially leading to misinterpretation of the images. In this paper, we propose a novel approach based on the discrete shape space of images (DSSI) to jointly reduce artifacts and noise in MR images. The proposed method restores MR images in multiple domains based on the distinct generation mechanisms of noise and artifacts. The images in multiple domains are analyzed in a non-Euclidean space. The DSSI is constructed as a Riemannian manifold to measure the intrinsic properties of images. Images are considered shapes from a geometric perspective, and the impact of similarity transformations (e.g., rotation, scaling, and translation) on image analysis is eliminated. The patch-based rank-ordered difference (PROD) detector is defined in k-space within the framework of DSSI to detect and remove sparse outliers that cause artifacts. In addition, a novel similarity function for images is defined using the DSSI and be used to design the improved filter. Finally, the convergence of the improved filter is theoretically analyzed, indicating that our method offers an effective estimator of the ideal image. The experimental results of various MR images demonstrate that the proposed approach outperforms classical and state-of-the-art methods for artifact correction and noise removal, both qualitatively and quantitatively. • We propose a method for enhancing MR image quality by reducing artifacts and noise simultaneously. Motivated by the properties of these undesired components, our method combines the advantages of both k-space and image space to enhance images. • We construct the discrete shape space of images (DSSI) to measure the intrinsic image similarity in Riemannian manifold by treating images as shapes from a geometric view. Therefore, image similarity is independent of similarity transformations. • The patch-based rank-ordered difference (PROD) is defined under the framework of the DSSI to detect and remove MR artifacts in k-space. The PROD is defined by a signed distance between patches as artifacts usually appear as sparse outliers with prominent features. • We design an improved filter using the similarity function of images defined in the DSSI to reduce noise in the image space. This is an interesting advantage when compared with the similarity function used in the classical NLM, which is rather sensitive to rotation. • We theoretically investigate the convergence of our improved filter and find that using the geometric information of images to measure the similarity between images can ensure that the restored image is an effective estimator of the ideal image. [ABSTRACT FROM AUTHOR]

• An new multi-task learning framework is devised for differential diagnosis of subjective cognitive decline and mild cognive impairment. • The proposed multi-task learning algorithm combines functional and structural brain information. • The proposed method can discover the most disease-related brain regions and their connectivity. • The extensive experiments demonstrate good classification performance against competing techniques. Early diagnosis and intervention of mild cognitive impairment (MCI) and its early stage (i.e., subjective cognitive decline (SCD)) is able to delay or reverse the disease progression. However, discrimination between SCD, MCI and healthy subjects accurately remains challenging. This paper proposes an auto-weighted centralised multi-task (AWCMT) learning framework for differential diagnosis of SCD and MCI. AWCMT is based on structural and functional connectivity information inferred from magnetic resonance imaging (MRI). To be specific, we devise a novel multi-task learning algorithm to combine neuroimaging functional and structural connective information. We construct a functional brain network through a sparse and low-rank machine learning method, and also a structural brain network via fibre bundle tracking. Those two networks are constructed separately and independently. Multi-task learning is then used to identify features integration of functional and structural connectivity. Hence, we can learn each task's significance automatically in a balanced way. By combining the functional and structural information, the most informative features of SCD and MCI are obtained for diagnosis. The extensive experiments on the public and self-collected datasets demonstrate that the proposed algorithm obtains better performance in classifying SCD, MCI and healthy people than traditional algorithms. The newly proposed method has good interpretability as it is able to discover the most disease-related brain regions and their connectivity. The results agree well with current clinical findings and provide new insights into early AD detection based on the multi-modal neuroimaging technique. [Display omitted] [ABSTRACT FROM AUTHOR]

A common feature of morphogenesis is the formation of three-dimensional structures from the folding of two-dimensional epithelial sheets, aided by cell shape changes at the cellular-level. Changes in cell shape must be studied in the context of cell-polarised biomechanical processes within the epithelial sheet. In epithelia with highly curved surfaces, finding single-cell alignment along a biological axis can be difficult to automate in silico. We present 'Origami', a MATLAB-based image analysis pipeline to compute direction-variant cell shape features along the epithelial apico-basal axis. Our automated method accurately computed direction vectors denoting the apico-basal axis in regions with opposing curvature in synthetic epithelia and fluorescence images of zebrafish embryos. As proof of concept, we identified different cell shape signatures in the developing zebrafish inner ear, where the epithelium deforms in opposite orientations to form different structures. Origami is designed to be user-friendly and is generally applicable to fluorescence images of curved epithelia. Author summary: During embryonic development, two-dimensional epithelial sheets bend and fold into complex three-dimensional structures–like paper in the origami art form. The genetic and biomechanical processes driving epithelial folding can be polarised in the epithelium, leading to asymmetric shape changes at the single cell level. Defects in such epithelial shaping have been linked to many developmental anomalies and diseases. It is, therefore, important not only to quantify shape change at the single cell level, but also to orientate these asymmetrical changes along an epithelial axis of polarity when studying morphogenetic processes. Origami is a MATLAB-based software that has been developed to automatically extract such single-cell asymmetrical shape features along the epithelial apico-basal axis from fluorescence microscopy images of folding epithelia. Origami provides a solution to computing directional vectors along the epithelial apico-basal axis followed by extracting direction-variant shape features of each segmented cell. It is generally applicable to epithelial structures regardless of complexity or direction of folding and is robust to imaging conditions. As proof of concept, Origami successfully differentiated between different cell shape signatures in highly curved structures at different developmental timepoints in the zebrafish inner ear. [ABSTRACT FROM AUTHOR]

Abstract: One of the important obstacles in the image-based analysis of the human face is the 3D nature of the problem and the 2D nature of most imaging systems used for biometric applications. Due to this, accuracy is strongly influenced by the viewpoint of the images, being frontal views the most thoroughly studied. However, when fully automatic face analysis systems are designed, capturing frontal-view images cannot be guaranteed. Examples of this situation can be found in surveillance systems, car driver images or whenever there are architectural constraints that prevent from placing a camera frontal to the subject. Taking advantage of the fact that most facial features lie approximately on the same plane, we propose the use of projective geometry across different views. An active shape model constructed with frontal-view images can then be directly applied to the segmentation of pictures taken from other viewpoints. The proposed extension demonstrates being significantly more invariant than the standard approach. Validation of the method is presented in 360 images from the AV@CAR database, systematically divided into three different rotations (to both sides), as well as upper and lower views due to nodding. The presented tests are among the largest quantitative results reported to date in face segmentation under varying poses. [Copyright &y& Elsevier]

Abstract: This article proposes an extension of Haar-like features for their use in rapid object detection systems. These features differ from the traditional ones in that their rectangles are assigned optimal weights so as to maximize their ability to discriminate objects from clutter (non-objects). These features maintain the simplicity of evaluation of the traditional formulation while being more discriminative. The proposed features were trained to detect two types of objects: human frontal faces and human heart regions. Our experimental results suggest that the object detectors based on the proposed features are more accurate and faster than the object detectors built with traditional Haar-like features. [Copyright &y& Elsevier]

Abstract: A new technique (SPASM) based on a 3D-ASM is presented for automatic segmentation of cardiac MRI image data sets consisting of multiple planes with arbitrary orientations, and with large undersampled regions. Model landmark positions are updated in a two-stage iterative process. First, landmark positions close to intersections with images are updated. Second, the update information is propagated to the regions without image information, such that new locations for the whole set of the model landmarks are obtained. Feature point detection is performed by a fuzzy inference system, based on fuzzy C-means clustering. Model parameters were optimized on a computer cluster and the computational load distributed by grid computing. SPASM was applied to image data sets with an increasing sparsity (from 2 to 11 slices) comprising images with different orientations and stemming from different MRI acquisition protocols. Segmentation outcomes and calculated volumes were compared to manual segmentation on a dense short-axis data configuration in a 3D manner. For all data configurations, (sub-)pixel accuracy was achieved. Performance differences between data configurations were significantly different (p <0.05) for SA data sets with less than 6 slices, but not clinically relevant (volume differences<4ml). Comparison to results from other 3D model-based methods showed that SPASM performs comparable to or better than these other methods, but SPASM uses considerably less image data. Sensitivity to initial model placement proved to be limited within a range of position perturbations of approximately 20mm in all directions. [Copyright &y& Elsevier]

In this paper, a new technique coined two-dimensional principal component analysis (2DPCA) is developed for image representation. As opposed to PCA, 2DPCA is based on 2D image matrices rather than 1D vectors so the image matrix does not need to be transformed into a vector prior to feature extraction. Instead, an image covariance matrix is constructed directly using the original image matrices, and its eigenvectors are derived for image feature extraction. To test 2DPCA and evaluate its performance, a series of experiments were performed on three face image databases: ORL, AR, and Yale face databases. The recognition rate across all trials was higher using 2DPCA than PCA. The experimental results also indicated that the extraction of image features is computationally more efficient using 2DPCA than PCA. [ABSTRACT FROM AUTHOR]

In this paper, a novel image projection analysis method (UIPDA) is first developed for image feature extraction. In contrast to Liu's projection discriminant method, UIPDA has the desirable property that the projected feature vectors are mutually uncorrelated. Also, a new LDA technique called EULDA is presented for further feature extraction. The proposed methods are tested on the ORL and the NUST603 face databases. The experimental results demonstrate that: (i) UIPDA is superior to Liu's projection discriminant method and more efficient than Eigenfaces and Fisherfaces; (ii) EULDA outperforms the existing PCA plus LDA strategy; (iii) UIPDA plus EULDA is a very effective two-stage strategy for image feature extraction. [ABSTRACT FROM AUTHOR]

In this paper, a Complex LDA based combined Fisherfaces framework, coined Complex Fisherfaces, is developed for face feature extraction and recognition. In this framework, Principal Component Analysis (PCA) and Kernel PCA (KPCA) are first used for feature extraction. Then, the resulting PCA-based linear features and KPCA-based nonlinear features are integrated by complex vectors and, Complex LDA is further employed for feature fusion. The proposed method is tested on a subset of FERET database. The experimental results demonstrate that Complex Fisherfaces outperforms Fisherfaces and Kernel Fisherfaces. Also, the complex vector based parallel feature fusion strategy is demonstrated to be much more effective and robust than the super-vector based serial feature fusion strategy for face recognition. [Copyright &y& Elsevier]

• We design a similarity-aware graph in GCN to consider disease statuses. • We propose an adaptive mechanism to estimate similarities. • We propose a calibration mechanism to fuse functional and structural information. • We obtain promising performance for disease deterioration prediction. Image, graphical abstract Graph convolution networks (GCN) have been successfully applied in disease prediction tasks as they capture interactions (i.e., edges and edge weights on the graph) between individual elements. The interactions in existing works are constructed by fusing similarity between imaging information and distance between non-imaging information, whereas disregarding the disease status of those individuals in the training set. Besides, the similarity is being evaluated by computing the correlation distance between feature vectors, which limits prediction performance, especially for predicting significant memory concern (SMC) and mild cognitive impairment (MCI). In this paper, we propose three mechanisms to improve GCN, namely similarity-aware adaptive calibrated GCN (SAC-GCN), for predicting SMC and MCI. First, we design a similarity-aware graph using different receptive fields to consider disease status. The labelled subjects on the graph are only connected with those labelled subjects with the same status. Second, we propose an adaptive mechanism to evaluate similarity. Specifically, we construct initial GCN with evaluating similarity by using traditional correlation distance, then pre-train the initial GCN by using training samples and use it to score all subjects. Then, the difference between these scores replaces correlation distance to update similarity. Last, we devise a calibration mechanism to fuse functional magnetic resonance imaging (fMRI) and diffusion tensor imaging (DTI) information into edges. The proposed method is tested on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Experimental results demonstrate that our proposed method is useful to predict disease-induced deterioration and superior to other related algorithms, with a mean classification accuracy of 86.83% in our prediction tasks. [ABSTRACT FROM AUTHOR]

Invasive coronary angiography remains the gold standard for diagnosing coronary artery disease, which may be complicated by both, patient-specific anatomy and image quality. Deep learning techniques aimed at detecting coronary artery stenoses may facilitate the diagnosis. However, previous studies have failed to achieve superior accuracy and performance for real-time labeling. Our study is aimed at confirming the feasibility of real-time coronary artery stenosis detection using deep learning methods. To reach this goal we trained and tested eight promising detectors based on different neural network architectures (MobileNet, ResNet-50, ResNet-101, Inception ResNet, NASNet) using clinical angiography data of 100 patients. Three neural networks have demonstrated superior results. The network based on Faster-RCNN Inception ResNet V2 is the most accurate and it achieved the mean Average Precision of 0.95, F1-score 0.96 and the slowest prediction rate of 3 fps on the validation subset. The relatively lightweight SSD MobileNet V2 network proved itself as the fastest one with a low mAP of 0.83, F1-score of 0.80 and a mean prediction rate of 38 fps. The model based on RFCN ResNet-101 V2 has demonstrated an optimal accuracy-to-speed ratio. Its mAP makes up 0.94, F1-score 0.96 while the prediction speed is 10 fps. The resultant performance-accuracy balance of the modern neural networks has confirmed the feasibility of real-time coronary artery stenosis detection supporting the decision-making process of the Heart Team interpreting coronary angiography findings. [ABSTRACT FROM AUTHOR]

[Display omitted] • A multiscale reference model is developed by combining CBCT and μCT scans. • Geodesic wave propagation is proposed for prescribing cortical bone orthotropy. • Variations in the mandible material properties are represented by 16 zones. • Muscles are realistically modeled using uniformly distributed muscle fiber bundles. • The range of dental problems can be studied by editing the reference intact model. Computer modeling of lower mandible physiology remains challenging because prescribing realistic material characteristics and boundary conditions from medical scans requires advanced equipment and skill sets. The objective of this study is to provide a framework that could reduce simplifications made and inconsistency (in terms of geometry, materials, and boundary conditions) among further studies on the topic. The OpenMandible framework offers: 1) the first publicly available multiscale model of the mandible developed by combining cone beam computerized tomography (CBCT) and μCT imaging modalities, and 2) a C++ software tool for the generation of simulation-ready models (tet4 and hex8 elements). In addition to the application of conventional (Neumann and Dirichlet) boundary conditions, OpenMandible introduces a novel geodesic wave propagation - based approach for incorporating orthotropic micromechanical characteristics of cortical bone, and a unique algorithm for modeling muscles as uniformly directed vectors. The base intact model includes the mandible (spongy and compact bone), 14 teeth (comprising dentin, enamel, periodontal ligament, and pulp), simplified temporomandibular joints, and masticatory muscles (masseter, temporalis, medial, and lateral pterygoid). The complete source code, executables, showcases, and sample data are freely available on the public repository: https://github.com/ArsoVukicevic/OpenMandible. It has been demonstrated that by slightly editing the baseline model, one can study different "virtual" treatments or diseases, including tooth restoration, placement of implants, mandible bone degradation, and others. OpenMandible eases the community to undertake a broad range of studies on the topic, while increasing their consistency and reproducibility. At the same time, the needs for dedicated equipment and skills for developing realistic simulation models are significantly reduced. [ABSTRACT FROM AUTHOR]

• An automatic IQA model for CMR images with different views is presented. • A generalised optimisation-based deep meta-learning model is proposed. • The model can detect respiratory and cardiac motions, aliasing and Gibbs artefacts. • This model is able to address the lack of large annotated datasets. Cardiovascular magnetic resonance (CMR) imaging is a powerful modality in functional and anatomical assessment for various cardiovascular diseases. Sufficient image quality is essential to achieve proper diagnosis and treatment. A large number of medical images, the variety of imaging artefacts, and the workload of imaging centres are amongst the factors that reveal the necessity of automatic image quality assessment (IQA). However, automated IQA requires access to bulk annotated datasets for training deep learning (DL) models. Labelling medical images is a tedious, costly and time-consuming process, which creates a fundamental challenge in proposing DL-based methods for medical applications. This study aims to present a new method for CMR IQA when there is limited access to annotated datasets. The proposed generalised deep meta-learning model can evaluate the quality by learning tasks in the prior stage and then fine-tuning the resulting model on a small labelled dataset of the desired tasks. This model was evaluated on the data of over 6,000 subjects from the UK Biobank for five defined tasks, including detecting respiratory motion, cardiac motion, Aliasing and Gibbs ringing artefacts and images without artefacts. The results of extensive experiments show the superiority of the proposed model. Besides, comparing the model's accuracy with the domain adaptation model indicates a significant difference by using only 64 annotated images related to the desired tasks. The proposed model can identify unknown artefacts in images with acceptable accuracy, which makes it suitable for medical applications and quality assessment of large cohorts. https://github.com/HosseinSimchi/META-IQA-CMRImages [ABSTRACT FROM AUTHOR]

• A network estimation method can automatically calibrate data quality and integrate modula prior. • A multi-task feature learning method is developed using multi-modal data. • A joint non-convex regularizer is designed for subspace learning. • Our method has achieved good automatic diagnosis and classification performance. Detection of early stages of Alzheimer's disease (AD) (i.e., mild cognitive impairment (MCI)) is important to maximize the chances to delay or prevent progression to AD. Brain connectivity networks inferred from medical imaging data have been commonly used to distinguish MCI patients from normal controls (NC). However, existing methods still suffer from limited performance, and classification remains mainly based on single modality data. This paper proposes a new model to automatically diagnosing MCI (early MCI (EMCI) and late MCI (LMCI)) and its earlier stages (i.e., significant memory concern (SMC)) by combining low-rank self-calibrated functional brain networks and structural brain networks for joint multi-task learning. Specifically, we first develop a new functional brain network estimation method. We introduce data quality indicators for self-calibration, which can improve data quality while completing brain network estimation, and perform correlation analysis combined with low-rank structure. Second, functional and structural connected neuroimaging patterns are integrated into our multi-task learning model to select discriminative and informative features for fine MCI analysis. Different modalities are best suited to undertake distinct classification tasks, and similarities and differences among multiple tasks are best determined through joint learning to determine most discriminative features. The learning process is completed by non-convex regularizer, which effectively reduces the penalty bias of trace norm and approximates the original rank minimization problem. Finally, the most relevant disease features classified using a support vector machine (SVM) for MCI identification. Experimental results show that our method achieves promising performance with high classification accuracy and can effectively discriminate between different sub-stages of MCI. Image, graphical abstract [ABSTRACT FROM AUTHOR]

Osteoporosis is an age-associated disease characterised by low bone mineral density (BMD) and micro-architectural deterioration leading to enhanced fracture risk. Conventional dual-energy X-ray absorptiometry (DXA) analysis has facilitated our understanding of BMD reduction in specific regions of interest (ROIs) within the femur, but cannot resolve spatial BMD patterns nor reflect age-related changes in bone microarchitecture due to its inherent averaging of pixel BMD values into large ROIs. To address these limitations and develop a comprehensive model of involutional bone loss, this paper presents a fully automatic pipeline to build a spatio-temporal atlas of ageing bone in the proximal femur. A new technique, termed DXA region free analysis (DXA RFA), is proposed to eliminate morphological variation between DXA scans by warping each image into a reference template. To construct the atlas, we use unprocessed DXA data from Caucasian women aged 20–97 years participating in three cohort studies in Western Europe (${n} >13$ ,000). A novel calibration procedure, termed quantile matching regression, is proposed to integrate data from different DXA manufacturers. Pixel-wise BMD evolution with ageing was modelled using smooth quantile curves. This technique enables characterisation of spatially-complex BMD change patterns with ageing, visualised using heat-maps. Furthermore, quantile curves plotted at different pixel coordinates showed consistently different rates of bone loss at different regions within the femoral neck. Given the close relationship between spatio-temporal bone loss and osteoporotic fracture, improved understanding of the bone ageing process could lead to enhanced prognostic, preventive and therapeutic strategies for the disease. [ABSTRACT FROM AUTHOR]

Recent works show that generative adversarial networks (GANs) can be successfully applied to image synthesis and semi-supervised learning, where, given a small labeled database and a large unlabeled database, the goal is to train a powerful classifier. In this paper, we trained a retinal image synthesizer and a semi-supervised learning method for automatic glaucoma assessment using an adversarial model on a small glaucoma-labeled database and a large unlabeled database. Various studies have shown that glaucoma can be monitored by analyzing the optic disc and its surroundings, and for that reason, the images used in this paper were automatically cropped around the optic disc. The novelty of this paper is to propose a new retinal image synthesizer and a semi-supervised learning method for glaucoma assessment based on the deep convolutional GANs. In addition, and to the best of our knowledge, this system is trained on an unprecedented number of publicly available images (86926 images). This system, hence, is not only able to generate images synthetically but to provide labels automatically. Synthetic images were qualitatively evaluated using t-SNE plots of features associated with the images and their anatomical consistency was estimated by measuring the proportion of pixels corresponding to the anatomical structures around the optic disc. The resulting image synthesizer is able to generate realistic (cropped) retinal images, and subsequently, the glaucoma classifier is able to classify them into glaucomatous and normal with high accuracy (AUC = 0.9017). The obtained retinal image synthesizer and the glaucoma classifier could then be used to generate an unlimited number of cropped retinal images with glaucoma labels. [ABSTRACT FROM AUTHOR]

Population imaging studies generate data for developing and implementing personalised health strategies to prevent, or more effectively treat disease. Large prospective epidemiological studies acquire imaging for pre-symptomatic populations. These studies enable the early discovery of alterations due to impending disease, and enable early identification of individuals at risk. Such studies pose new challenges requiring automatic image analysis. To date, few large-scale population-level cardiac imaging studies have been conducted. One such study stands out for its sheer size, careful implementation, and availability of top quality expert annotation; the UK Biobank (UKB). The resulting massive imaging datasets (targeting ca. 100,000 subjects) has put published approaches for cardiac image quantification to the test. In this paper, we present and evaluate a cardiac magnetic resonance (CMR) image analysis pipeline that properly scales up and can provide a fully automatic analysis of the UKB CMR study. Without manual user interactions, our pipeline performs end-to-end image analytics from multi-view cine CMR images all the way to anatomical and functional bi-ventricular quantification. All this, while maintaining relevant quality controls of the CMR input images, and resulting image segmentations. To the best of our knowledge, this is the first published attempt to fully automate the extraction of global and regional reference ranges of all key functional cardiovascular indexes, from both left and right cardiac ventricles, for a population of 20,000 subjects imaged at 50 time frames per subject, for a total of one million CMR volumes. In addition, our pipeline provides 3D anatomical bi-ventricular models of the heart. These models enable the extraction of detailed information of the morphodynamics of the two ventricles for subsequent association to genetic, omics, lifestyle habits, exposure information, and other information provided in population imaging studies. We validated our proposed CMR analytics pipeline against manual expert readings on a reference cohort of 4620 subjects with contour delineations and corresponding clinical indexes. Our results show broad significant agreement between the manually obtained reference indexes, and those automatically computed via our framework. 80.67% of subjects were processed with mean contour distance of less than 1 pixel, and 17.50% with mean contour distance between 1 and 2 pixels. Finally, we compare our pipeline with a recently published approach reporting on UKB data, and based on deep learning. Our comparison shows similar performance in terms of segmentation accuracy with respect to human experts. [ABSTRACT FROM AUTHOR]