Publications
166 results found
Tarroni G, Bai W, Oktay O, et al., 2020, Large-scale quality control of cardiac imaging in population studies: application to UK Biobank, Scientific Reports, Vol: 10, ISSN: 2045-2322
In large population studies such as the UK Biobank (UKBB), quality control of the acquired images by visual assessment isunfeasible. In this paper, we apply a recently developed fully-automated quality control pipeline for cardiac MR (CMR) imagesto the first 19,265 short-axis (SA) cine stacks from the UKBB. We present the results for the three estimated quality metrics(heart coverage, inter-slice motion and image contrast in the cardiac region) as well as their potential associations with factorsincluding acquisition details and subject-related phenotypes. Up to 14.2% of the analysed SA stacks had sub-optimal coverage(i.e. missing basal and/or apical slices), however most of them were limited to the first year of acquisition. Up to 16% of thestacks were affected by noticeable inter-slice motion (i.e. average inter-slice misalignment greater than 3.4 mm). Inter-slicemotion was positively correlated with weight and body surface area. Only 2.1% of the stacks had an average end-diastoliccardiac image contrast below 30% of the dynamic range. These findings will be highly valuable for both the scientists involvedin UKBB CMR acquisition and for the ones who use the dataset for research purposes.
Chen C, Ouyang C, Tarroni G, et al., 2020, Unsupervised multi-modal style transfer for cardiac MR segmentation, MICCAI STACOM Workshop, Publisher: Springer International Publishing, Pages: 209-219, ISSN: 0302-9743
In this work, we present a fully automatic method to segment cardiac structures from late-gadolinium enhanced (LGE) images without using labelled LGE data for training, but instead by transferring the anatomical knowledge and features learned on annotated balanced steady-state free precession (bSSFP) images, which are easier to acquire. Our framework mainly consists of two neural networks: a multi-modal image translation network for style transfer and a cascaded segmentation network for image segmentation. The multi-modal image translation network generates realistic and diverse synthetic LGE images conditioned on a single annotated bSSFP image, forming a synthetic LGE training set. This set is then utilized to fine-tune the segmentation network pre-trained on labelled bSSFP images, achieving the goal of unsupervised LGE image segmentation. In particular, the proposed cascaded segmentation network is able to produce accurate segmentation by taking both shape prior and image appearance into account, achieving an average Dice score of 0.92 for the left ventricle, 0.83 for the myocardium, and 0.88 for the right ventricle on the test set.
Biffi C, Doumou G, Duan J, et al., 2020, Explainable anatomical shape analysis through deep hierarchical generative models., Publisher: arXiv
Quantification of anatomical shape changes currently relies on scalar global indexes which are largely insensitive to regional or asymmetric modifications. Accurate assessment of pathology-driven anatomical remodeling is a crucial step for the diagnosis and treatment of many conditions. Deep learning approaches have recently achieved wide success in the analysis of medical images, but they lack interpretability in the feature extraction and decision processes. In this work, we propose a new interpretable deep learning model for shape analysis. In particular, we exploit deep generative networks to model a population of anatomical segmentations through a hierarchy of conditional latent variables. At the highest level of this hierarchy, a two-dimensional latent space is simultaneously optimised to discriminate distinct clinical conditions, enabling the direct visualisation of the classification space. Moreover, the anatomical variability encoded by this discriminative latent space can be visualised in the segmentation space thanks to the generative properties of the model, making the classification task transparent. This approach yielded high accuracy in the categorisation of healthy and remodelled left ventricles when tested on unseen segmentations from our own multi-centre dataset as well as in an external validation set, and on hippocampi from healthy controls and patients with Alzheimer's disease when tested on ADNI data. More importantly, it enabled the visualisation in three-dimensions of both global and regional anatomical features which better discriminate between the conditions under exam. The proposed approach scales effectively to large populations, facilitating highthroughput analysis of normal anatomy and pathology in largescale studies of volumetric imaging.
Ruijsink B, Puyol-Antón E, Li Y, et al., 2020, Quality-aware semi-supervised learning for CMR segmentation., Pages: 97-107
One of the challenges in developing deep learning algorithms for medical image segmentation is the scarcity of annotated training data. To overcome this limitation, data augmentation and semi-supervised learning (SSL) methods have been developed. However, these methods have limited effectiveness as they either exploit the existing data set only (data augmentation) or risk negative impact by adding poor training examples (SSL). Segmentations are rarely the final product of medical image analysis -they are typically used in downstream tasks to infer higher-order patterns to evaluate diseases. Clinicians take into account a wealth of prior knowledge on biophysics and physiology when evaluating image analysis results. We have used these clinical assessments in previous works to create robust quality-control (QC) classifiers for automated cardiac magnetic resonance (CMR) analysis. In this paper, we propose a novel scheme that uses QC of the downstream task to identify high quality outputs of CMR segmentation networks, that are subsequently utilised for further network training. In essence, this provides quality-aware augmentation of training data in a variant of SSL for segmentation networks (semiQCSeg). We evaluate our approach in two CMR segmentation tasks (aortic and short axis cardiac volume segmentation) using UK Biobank data and two commonly used network architectures (U-net and a Fully Convolutional Network) and compare against supervised and SSL strategies. We show that semiQCSeg improves training of the segmentation networks. It decreases the need for labelled data, while outperforming the other methods in terms of Dice and clinical metrics. SemiQCSeg can be an efficient approach for training segmentation networks for medical image data when labelled datasets are scarce.
Lu P, Qiu H, Qin C, et al., 2020, Going Deeper into Cardiac Motion Analysis to Model Fine Spatio-Temporal Features, 24th Conference on Medical Image Understanding and Analysis (MIUA), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 294-306, ISSN: 1865-0929
- Author Web Link
- Cite
- Citations: 5
Suzuki H, Venkataraman AV, Bai W, et al., 2019, Associations of regional brain structural differences with aging, modifiable risk factors for dementia, and cognitive performance, JAMA Network Open, Vol: 2, Pages: 1-19, ISSN: 2574-3805
Importance Identifying brain regions associated with risk factors for dementia could guide mechanistic understanding of risk factors associated with Alzheimer disease (AD).Objectives To characterize volume changes in brain regions associated with aging and modifiable risk factors for dementia (MRFD) and to test whether volume differences in these regions are associated with cognitive performance.Design, Setting, and Participants This cross-sectional study used data from UK Biobank participants who underwent T1-weighted structural brain imaging from August 5, 2014, to October 14, 2016. A voxelwise linear model was applied to test for regional gray matter volume differences associated with aging and MRFD (ie, hypertension, diabetes, obesity, and frequent alcohol use). The potential clinical relevance of these associations was explored by comparing their neuroanatomical distributions with the regional brain atrophy found with AD. Mediation models for risk factors, brain volume differences, and cognitive measures were tested. The primary hypothesis was that common, overlapping regions would be found. Primary analysis was conducted on April 1, 2018.Main Outcomes and Measures Gray matter regions that showed relative atrophy associated with AD, aging, and greater numbers of MRFD.Results Among 8312 participants (mean [SD] age, 62.4 [7.4] years; 3959 [47.1%] men), aging and 4 major MRFD (ie, hypertension, diabetes, obesity, and frequent alcohol use) had independent negative associations with specific gray matter volumes. These regions overlapped neuroanatomically with those showing lower volumes in participants with AD, including the posterior cingulate cortex, the thalamus, the hippocampus, and the orbitofrontal cortex. Associations between these MRFD and spatial memory were mediated by differences in posterior cingulate cortex volume (β = 0.0014; SE = 0.0006; P = .02).Conclusions and Relevance This cross-sectional study
Halliday BP, Balaban G, Costa CM, et al., 2019, Improving Arrhythmic Risk Stratification in Non-Ischemic Dilated Cardiomyopathy Through the Evaluation of Novel Scar Characteristics Using CMR, Scientific Sessions of the American-Heart-Association, Publisher: LIPPINCOTT WILLIAMS & WILKINS, ISSN: 0009-7322
Chen C, Biffi C, Tarroni G, et al., 2019, Learning shape priors for robust cardiac MR segmentation from multi-view images, International Conference on Medical Image Computing and Computer-Assisted Intervention, Publisher: Springer International Publishing, Pages: 523-531, ISSN: 0302-9743
Cardiac MR image segmentation is essential for the morphological and functional analysis of the heart. Inspired by how experienced clinicians assess the cardiac morphology and function across multiple standard views (i.e. long- and short-axis views), we propose a novel approach which learns anatomical shape priors across different 2D standard views and leverages these priors to segment the left ventricular (LV) myocardium from short-axis MR image stacks. The proposed segmentation method has the advantage of being a 2D network but at the same time incorporates spatial context from multiple, complementary views that span a 3D space. Our method achieves accurate and robust segmentation of the myocardium across different short-axis slices (from apex to base), outperforming baseline models (e.g. 2D U-Net, 3D U-Net) while achieving higher data efficiency. Compared to the 2D U-Net, the proposed method reduces the mean Hausdorff distance (mm) from 3.24 to 2.49 on the apical slices, from 2.34 to 2.09 on the middle slices and from 3.62 to 2.76 on the basal slices on the test set, when only 10% of the training data was used.
Bhuva AN, Bai W, Lau C, et al., 2019, A multicenter, scan-rescan, human and machine learning CMR study to test generalizability and precision in imaging biomarker analysis, Circulation: Cardiovascular Imaging, Vol: 12, Pages: 1-11, ISSN: 1941-9651
Background:Automated analysis of cardiac structure and function using machine learning (ML) has great potential, but is currently hindered by poor generalizability. Comparison is traditionally against clinicians as a reference, ignoring inherent human inter- and intraobserver error, and ensuring that ML cannot demonstrate superiority. Measuring precision (scan:rescan reproducibility) addresses this. We compared precision of ML and humans using a multicenter, multi-disease, scan:rescan cardiovascular magnetic resonance data set.Methods:One hundred ten patients (5 disease categories, 5 institutions, 2 scanner manufacturers, and 2 field strengths) underwent scan:rescan cardiovascular magnetic resonance (96% within one week). After identification of the most precise human technique, left ventricular chamber volumes, mass, and ejection fraction were measured by an expert, a trained junior clinician, and a fully automated convolutional neural network trained on 599 independent multicenter disease cases. Scan:rescan coefficient of variation and 1000 bootstrapped 95% CIs were calculated and compared using mixed linear effects models.Results:Clinicians can be confident in detecting a 9% change in left ventricular ejection fraction, with greater than half of coefficient of variation attributable to intraobserver variation. Expert, trained junior, and automated scan:rescan precision were similar (for left ventricular ejection fraction, coefficient of variation 6.1 [5.2%–7.1%], P=0.2581; 8.3 [5.6%–10.3%], P=0.3653; 8.8 [6.1%–11.1%], P=0.8620). Automated analysis was 186× faster than humans (0.07 versus 13 minutes).Conclusions:Automated ML analysis is faster with similar precision to the most precise human techniques, even when challenged with real-world scan:rescan data. Assessment of multicenter, multi-vendor, multi-field strength scan:rescan data (available at www.thevolumesresource.com) permits a generalizable assessment of ML precision and may facili
Balaban G, Halliday BP, Bai W, et al., 2019, Scar shape analysis and simulated electrical instabilities in a non-ischemic dilated cardiomyopathy patient cohort., PLoS Computational Biology, Vol: 15, Pages: 1-18, ISSN: 1553-734X
This paper presents a morphological analysis of fibrotic scarring in non-ischemic dilated cardiomyopathy, and its relationship to electrical instabilities which underlie reentrant arrhythmias. Two dimensional electrophysiological simulation models were constructed from a set of 699 late gadolinium enhanced cardiac magnetic resonance images originating from 157 patients. Areas of late gadolinium enhancement (LGE) in each image were assigned one of 10 possible microstructures, which modelled the details of fibrotic scarring an order of magnitude below the MRI scan resolution. A simulated programmed electrical stimulation protocol tested each model for the possibility of generating either a transmural block or a transmural reentry. The outcomes of the simulations were compared against morphological LGE features extracted from the images. Models which blocked or reentered, grouped by microstructure, were significantly different from one another in myocardial-LGE interface length, number of components and entropy, but not in relative area and transmurality. With an unknown microstructure, transmurality alone was the best predictor of block, whereas a combination of interface length, transmurality and number of components was the best predictor of reentry in linear discriminant analysis.
Wang S, Dai C, Mo Y, et al., 2019, Automatic Brain Tumour Segmentation and Biophysics-Guided Survival Prediction, MICCAI BraTS 2019 Challenge
Gliomas are the most common malignant brain tumourswith intrinsicheterogeneity. Accurate segmentation of gliomas and theirsub-regions onmulti-parametric magnetic resonance images (mpMRI)is of great clinicalimportance, which defines tumour size, shape andappearance and providesabundant information for preoperative diag-nosis, treatment planning andsurvival prediction. Recent developmentson deep learning have significantlyimproved the performance of auto-mated medical image segmentation. In thispaper, we compare severalstate-of-the-art convolutional neural network modelsfor brain tumourimage segmentation. Based on the ensembled segmentation, wepresenta biophysics-guided prognostic model for patient overall survivalpredic-tion which outperforms a data-driven radiomics approach. Our methodwonthe second place of the MICCAI 2019 BraTS Challenge for theoverall survivalprediction.
Duan J, Schlemper J, Qin C, et al., 2019, VS-Net: variable splitting network for accelerated parallel MRI reconstruction, International Conference on Medical Image Computing and Computer-Assisted Intervention, Publisher: Springer International Publishing, Pages: 713-722, ISSN: 0302-9743
In this work, we propose a deep learning approach for parallel magnetic resonance imaging (MRI) reconstruction, termed a variable splitting network (VS-Net), for an efficient, high-quality reconstruction of undersampled multi-coil MR data. We formulate the generalized parallel compressed sensing reconstruction as an energy minimization problem, for which a variable splitting optimization method is derived. Based on this formulation we propose a novel, end-to-end trainable deep neural network architecture by unrolling the resulting iterative process of such variable splitting scheme. VS-Net is evaluated on complex valued multi-coil knee images for 4-fold and 6-fold acceleration factors. We show that VS-Net outperforms state-of-the-art deep learning reconstruction algorithms, in terms of reconstruction accuracy and perceptual quality. Our code is publicly available at https://github.com/j-duan/VS-Net.
Qureshi N, Kim S, Cantwell C, et al., 2019, Voltage during atrial fibrillation is superior to voltage during sinus rhythm in localizing areas of delayed enhancement on magnetic resonance imaging: An assessment of the posterior left atrium in patients with persistent atrial fibrillation, Heart Rhythm, Vol: 16, Pages: 1357-1367, ISSN: 1547-5271
BackgroundBipolar electrogram voltage during sinus rhythm (VSR) has been used as a surrogate for atrial fibrosis in guiding catheter ablation of persistent AF, but the fixed rate and wavefront characteristics present during sinus rhythm may not accurately reflect underlying functional vulnerabilities responsible for AF maintenance.ObjectivesWe hypothesized that given adequate temporal sampling, the spatial distribution of mean AF voltage (VmAF) should better correlate with delayed-enhancement MRI (MRI-DE) detected atrial fibrosis than VSR.MethodsAF was mapped (8s) during index ablation for persistent AF (20 patients) using a 20-pole catheter (660±28 points/map). Following cardioversion, VSR was mapped (557±326 points/map). Electroanatomic and MRI-DE maps were co-registered in 14 patients.Results(i) The time course of VmAF was assessed from 1-40 AF-cycles (∼8s) at 1113 locations. VmAF stabilized with sampling >4s (mean voltage error=0.05mV). (ii) Paired point analysis of VmAF from segments acquired 30s apart (3,667-sites, 15-patients), showed strong correlation (r=0.95, p<0.001). (iii) Delayed-enhancement (DE) was assessed across the posterior left atrial (LA) wall, occupying 33±13%. VmAF distributions (median[IQR]) were 0.21[0.14-0.35]mV in DE vs. 0.52[0.34-0.77]mV in Non-DE regions. VSR distributions were 1.34[0.65-2.48]mV in DE vs. 2.37[1.27-3.97]mV in Non-DE. A VmAF threshold of 0.35mV yielded sensitivity/specificity 75%/79% in detecting MRI-DE, compared with 63%/67% for VSR (1.8mV threshold).ConclusionThe correlation between low-voltage and posterior LA MRI-DE is significantly improved when acquired during AF vs. sinus rhythm. With adequate sampling, mean AF voltage is a reproducible marker reflecting the functional response to the underlying persistent AF substrate.
Duan J, Bello G, Schlemper J, et al., 2019, Automatic 3D bi-ventricular segmentation of cardiac images by a shape-refined multi-task deep learning approach, IEEE Transactions on Medical Imaging, Vol: 38, Pages: 2151-2164, ISSN: 0278-0062
Deep learning approaches have achieved state-of-the-art performance incardiac magnetic resonance (CMR) image segmentation. However, most approaches have focused on learning image intensity features for segmentation, whereas the incorporation of anatomical shape priors has received less attention. In this paper, we combine a multi-task deep learning approach with atlas propagation to develop a shape-constrained bi-ventricular segmentation pipeline for short-axis CMR volumetric images. The pipeline first employs a fully convolutional network (FCN) that learns segmentation and landmark localisation tasks simultaneously. The architecture of the proposed FCN uses a 2.5D representation, thus combining the computational advantage of 2D FCNs networks and the capability of addressing 3D spatial consistency without compromising segmentation accuracy. Moreover, the refinement step is designed to explicitly enforce a shape constraint and improve segmentation quality. This step is effective for overcoming image artefacts (e.g. due to different breath-hold positions and large slice thickness), which preclude the creation of anatomically meaningful 3D cardiac shapes. The proposed pipeline is fully automated, due to network's ability to infer landmarks, which are then used downstream in the pipeline to initialise atlas propagation. We validate the pipeline on 1831 healthy subjects and 649 subjects with pulmonary hypertension. Extensive numerical experiments on the two datasets demonstrate that our proposed method is robust and capable of producing accurate, high-resolution and anatomically smooth bi-ventricular3D models, despite the artefacts in input CMR volumes.
Dai C, Mo Y, Angelini E, et al., 2019, Transfer learning from partial annotations for whole brain segmentation, International Workshop on Medical Image Learning with Less Labels and Imperfect Data
Brain MR image segmentation is a key task in neuroimaging studies. It is commonly conducted using standard computational tools, such as FSL, SPM, multi-atlas segmentation etc, which are often registration-based and suffer from expensive computation cost. Recently, there is an increased interest using deep neural networks for brain image segmentation, which have demonstrated advantages in both speed and performance. However, neural networks-based approaches normally require a large amount of manual annotations for optimising the massive amount of network parameters. For 3D networks used in volumetric image segmentation, this has become a particular challenge, as a 3D network consists of many more parameters compared to its 2D counterpart. Manual annotation of 3D brain images is extremely time-consuming and requires extensive involvement of trained experts. To address the challenge with limited manual annotations, here we propose a novel multi-task learning framework for brain image segmentation, which utilises a large amount of automatically generated partial annotations together with a small set of manually created full annotations for network training. Our method yields a high performance comparable to state-of-the-art methods for whole brain segmentation.
Bai W, Chen C, Tarroni G, et al., 2019, Self-supervised learning for cardiac MR image segmentation by anatomicalposition prediction, International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)
In the recent years, convolutional neural networks have transformed the field of medical image analysis due to their capacity to learn discriminative image features for a variety of classification and regression tasks. However, successfully learning these features requires a large amount of manuallyannotated data, which is expensive to acquire and limited by the availableresources of expert image analysts. Therefore, unsupervised, weakly-supervised and self-supervised feature learning techniques receive a lot of attention, which aim to utilise the vast amount of available data, while at the same time avoid or substantially reduce the effort of manual annotation. In this paper, we propose a novel way for training a cardiac MR image segmentation network, in which features are learnt in a self-supervised manner by predicting anatomical positions. The anatomical positions serve as a supervisory signal and do not require extra manual annotation. We demonstrate that this seemingly simple task provides a strong signal for feature learning and with self-supervised learning, we achieve a high segmentation accuracy that is better than or comparable to a U-net trained from scratch, especially at a small data setting. When only five annotated subjects are available, the proposed method improves the mean Dice metric from 0.811 to 0.852 for short-axis image segmentation, compared to the baseline U-net.
Chen C, Bai W, Davies RH, et al., 2019, Improving the generalizability of convolutional neural network-based segmentation on CMR images, Publisher: arXiv
Convolutional neural network (CNN) based segmentation methods provide anefficient and automated way for clinicians to assess the structure and functionof the heart in cardiac MR images. While CNNs can generally perform thesegmentation tasks with high accuracy when training and test images come fromthe same domain (e.g. same scanner or site), their performance often degradesdramatically on images from different scanners or clinical sites. We propose asimple yet effective way for improving the network generalization ability bycarefully designing data normalization and augmentation strategies toaccommodate common scenarios in multi-site, multi-scanner clinical imaging datasets. We demonstrate that a neural network trained on a single-sitesingle-scanner dataset from the UK Biobank can be successfully applied tosegmenting cardiac MR images across different sites and different scannerswithout substantial loss of accuracy. Specifically, the method was trained on alarge set of 3,975 subjects from the UK Biobank. It was then directly tested on600 different subjects from the UK Biobank for intra-domain testing and twoother sets for cross-domain testing: the ACDC dataset (100 subjects, 1 site, 2scanners) and the BSCMR-AS dataset (599 subjects, 6 sites, 9 scanners). Theproposed method produces promising segmentation results on the UK Biobank testset which are comparable to previously reported values in the literature, whilealso performing well on cross-domain test sets, achieving a mean Dice metric of0.90 for the left ventricle, 0.81 for the myocardium and 0.82 for the rightventricle on the ACDC dataset; and 0.89 for the left ventricle, 0.83 for themyocardium on the BSCMR-AS dataset. The proposed method offers a potentialsolution to improve CNN-based model generalizability for the cross-scanner andcross-site cardiac MR image segmentation task.
Bhuva A, Bai W, Lau C, et al., 2019, Fully automated left ventricular analysis matches clinician precision: a multi-centre, multi-vendor, multi-field strength, multi-disease scan:rescan CMR study, Publisher: OXFORD UNIV PRESS, Pages: 255-256, ISSN: 2047-2404
Tarroni G, Oktay O, Bai W, et al., 2019, Learning-based quality control for cardiac MR images, IEEE Transactions on Medical Imaging, Vol: 38, Pages: 1127-1138, ISSN: 0278-0062
The effectiveness of a cardiovascular magnetic resonance (CMR) scan depends on the ability of the operator to correctly tune the acquisition parameters to the subject being scanned and on the potential occurrence of imaging artefacts such as cardiac and respiratory motion. In the clinical practice, a quality control step is performed by visual assessment of the acquired images: however, this procedure is strongly operatordependent, cumbersome and sometimes incompatible with the time constraints in clinical settings and large-scale studies. We propose a fast, fully-automated, learning-based quality control pipeline for CMR images, specifically for short-axis image stacks. Our pipeline performs three important quality checks: 1) heart coverage estimation, 2) inter-slice motion detection, 3) image contrast estimation in the cardiac region. The pipeline uses a hybrid decision forest method - integrating both regression and structured classification models - to extract landmarks as well as probabilistic segmentation maps from both long- and short-axis images as a basis to perform the quality checks. The technique was tested on up to 3000 cases from the UK Biobank as well as on 100 cases from the UK Digital Heart Project, and validated against manual annotations and visual inspections performed by expert interpreters. The results show the capability of the proposed pipeline to correctly detect incomplete or corrupted scans (e.g. on UK Biobank, sensitivity and specificity respectively 88% and 99% for heart coverage estimation, 85% and 95% for motion detection), allowing their exclusion from the analysed dataset or the triggering of a new acquisition.
Cox DJ, Bai W, Price AN, et al., 2019, Ventricular remodeling in preterm infants: computational cardiac magnetic resonance atlasing shows significant early remodeling of the left ventricle, PEDIATRIC RESEARCH, Vol: 85, Pages: 807-815, ISSN: 0031-3998
- Author Web Link
- Cite
- Citations: 28
Robinson R, Valindria VV, Bai W, et al., 2019, Automated quality control in image segmentation: application to the UK Biobank cardiac MR imaging study, Journal of Cardiovascular Magnetic Resonance, Vol: 21, ISSN: 1097-6647
Background: The trend towards large-scale studies including population imaging poses new challenges in terms of quality control (QC). This is a particular issue when automatic processing tools, e.g. image segmentation methods, are employed to derive quantitative measures or biomarkers for later analyses. Manual inspection and visual QC of each segmentation isn't feasible at large scale. However, it's important to be able to automatically detect when a segmentation method fails so as to avoid inclusion of wrong measurements into subsequent analyses which could lead to incorrect conclusions. Methods: To overcome this challenge, we explore an approach for predicting segmentation quality based on Reverse Classification Accuracy, which enables us to discriminate between successful and failed segmentations on a per-cases basis. We validate this approach on a new, large-scale manually-annotated set of 4,800 cardiac magnetic resonance scans. We then apply our method to a large cohort of 7,250 cardiac MRI on which we have performed manual QC. Results: We report results used for predicting segmentation quality metrics including Dice Similarity Coefficient (DSC) and surface-distance measures. As initial validation, we present data for 400 scans demonstrating 99% accuracy for classifying low and high quality segmentations using predicted DSC scores. As further validation we show high correlation between real and predicted scores and 95% classification accuracy on 4,800 scans for which manual segmentations were available. We mimic real-world application of the method on 7,250 cardiac MRI where we show good agreement between predicted quality metrics and manual visual QC scores. Conclusions: We show that RCA has the potential for accurate and fully automatic segmentation QC on a per-case basis in the context of large-scale population imaging as in the UK Biobank Imaging Study.
Chen C, Bai W, Rueckert D, 2019, Multi-task learning for left atrial segmentation on GE-MRI, International Workshop on Statistical Atlases and Computational Models of the Heart, Publisher: Springer Verlag, Pages: 292-301, ISSN: 0302-9743
Segmentation of the left atrium (LA) is crucial for assessing its anatomy in both pre-operative atrial fibrillation (AF) ablation planning and post-operative follow-up studies. In this paper, we present a fully automated framework for left atrial segmentation in gadolinium-enhanced magnetic resonance images (GE-MRI) based on deep learning. We propose a fully convolutional neural network and explore the benefits of multi-task learning for performing both atrial segmentation and pre/post ablation classification. Our results show that, by sharing features between related tasks, the network can gain additional anatomical information and achieve more accurate atrial segmentation, leading to a mean Dice score of 0.901 on a test set of 20 3D MRI images. Code of our proposed algorithm is available at https://github.com/cherise215/atria_segmentation_2018/.
Gilbert K, Bai W, Mauger C, et al., 2019, Independent left ventricular morphometric atlases show consistent relationships with cardiovascular risk factors: A UK Biobank study, Scientific Reports, Vol: 9, ISSN: 2045-2322
Left ventricular (LV) mass and volume are important indicators of clinical and pre-clinical disease processes. However, much of the shape information present in modern imaging examinations is currently ignored. Morphometric atlases enable precise quantification of shape and function, but there has been no objective comparison of different atlases in the same cohort. We compared two independent LV atlases using MRI scans of 4547 UK Biobank participants: (i) a volume atlas derived by automatic non-rigid registration of image volumes to a common template, and (ii) a surface atlas derived from manually drawn epicardial and endocardial surface contours. The strength of associations between atlas principal components and cardiovascular risk factors (smoking, diabetes, high blood pressure, high cholesterol and angina) were quantified with logistic regression models and five-fold cross validation, using area under the ROC curve (AUC) and Akaike Information Criterion (AIC) metrics. Both atlases exhibited similar principal components, showed similar relationships with risk factors, and had stronger associations (higher AUC and lower AIC) than a reference model based on LV mass and volume, for all risk factors (DeLong p < 0.05). Morphometric variations associated with each risk factor could be quantified and visualized and were similar between atlases. UK Biobank LV shape atlases are robust to construction method and show stronger relationships with cardiovascular risk factors than mass and volume.
Oksuz I, Clough J, Bai W, et al., 2019, High-quality segmentation of low quality cardiac MR images using k-space artefact correction, Pages: 380-389
Deep learning methods have shown great success in segmenting the anatomical and pathological structures in medical images. This success is closely bounded with the quality of the images in the dataset that are being segmented. A commonly overlooked issue in the medical image analysis community is the vast amount of clinical images that have severe image artefacts. In this paper, we discuss the implications of image artefacts on cardiac MR segmentation and compare a variety of approaches for motion artefact correction with our proposed method Automap-GAN. Our method is based on the recently developed Automap reconstruction method, which directly reconstructs high quality MR images from k-space using deep learning. We propose to use a loss function that combines mean square error with structural similarity index to robustly segment poor-quality images. We train the reconstruction network to automatically correct for motion-related artefacts using synthetically corrupted CMR k-space data and uncorrected reconstructed images. In the experiments, we apply the proposed method to correct for motion artefacts on a large dataset of 1,400 subjects to improve image quality. The improvement of image quality is quantitatively assessed using segmentation accuracy as a metric. The segmentation is improved from 0.63 to 0.72 dice overlap after artefact correction. We quantitatively compare our method with a variety of techniques for recovering image quality to showcase the influence on segmentation. In addition, we qualitatively evaluate the proposed technique using k-space data containing real motion artefacts.
Balaban G, Halliday BP, Costa CM, et al., 2018, Fibrosis Microstructure Modulates Reentry in Non-ischemic Dilated Cardiomyopathy: Insights From Imaged Guided 2D Computational Modeling, Frontiers in Physiology, Vol: 9, ISSN: 1664-042X
Aims: Patients who present with non-ischemic dilated cardiomyopathy (NIDCM) andenhancement on late gadolinium magnetic resonance imaging (LGE-CMR), are at highrisk of sudden cardiac death (SCD). Further risk stratification of these patients basedon LGE-CMR may be improved through better understanding of fibrosis microstructure.Our aim is to examine variations in fibrosis microstructure based on LGE imaging, andquantify the effect on reentry inducibility and mechanism. Furthermore, we examine therelationship between transmural activation time differences and reentry.Methods and Results: 2D Computational models were created from a single short axisLGE-CMR image, with 401 variations in fibrosis type (interstitial, replacement) and density,as well as presence or absence of reduced conductivity (RC). Transmural activationtimes (TAT) were measured, as well as reentry incidence and mechanism. Reentrieswere inducible above specific density thresholds (0.8, 0.6 for interstitial, replacementfibrosis). RC reduced these thresholds (0.3, 0.4 for interstitial, replacement fibrosis) andincreased reentry incidence (48 no RC vs. 133 with RC). Reentries were classified as rotor,micro-reentry, or macro-reentry and depended on fibrosis micro-structure. Differencesin TAT at coupling intervals 210 and 500ms predicted reentry in the models (sensitivity89%, specificity 93%). A sensitivity analysis of TAT and reentry incidence showed thatthese quantities were robust to small changes in the pacing location.Conclusion: Computational models of fibrosis micro-structure underlying areas ofLGE in NIDCM provide insight into the mechanisms and inducibility of reentry, andtheir dependence upon the type and density of fibrosis. Transmural activation times,measured at the central extent of the scar, can potentially differentiate microstructureswhich support reentry.
Duan J, Schlemper J, Bai W, et al., 2018, Combining deep learning and shape priors for bi-ventricular segmentation of volumetric cardiac magnetic resonance images, MICCAI ShapeMI Workshop, Publisher: Springer Verlag, Pages: 258-267, ISSN: 0302-9743
In this paper, we combine a network-based method with image registration to develop a shape-based bi-ventricular segmentation tool for short-axis cardiac magnetic resonance (CMR) volumetric images. The method first employs a fully convolutional network (FCN) to learn the segmentation task from manually labelled ground truth CMR volumes. However, due to the presence of image artefacts in the training dataset, the resulting FCN segmentation results are often imperfect. As such, we propose a second step to refine the FCN segmentation. This step involves performing a non-rigid registration with multiple high-resolution bi-ventricular atlases, allowing the explicit shape priors to be inferred. We validate the proposed approach on 1831 healthy subjects and 200 subjects with pulmonary hypertension. Numerical experiments on the two datasets demonstrate that our approach is capable of producing accurate, high-resolution and anatomically smooth bi-ventricular models, despite the artefacts in the input CMR volumes.
Sinclair M, Baumgartner CF, Matthew J, et al., 2018, Human-level performance on automatic head biometrics in fetal ultrasound using fully convolutional neural networks, International Engineering in Medicine and Biology Conference, Pages: 714-717
Measurement of head biometrics from fetal ultrasonography images is of keyimportance in monitoring the healthy development of fetuses. However, theaccurate measurement of relevant anatomical structures is subject to largeinter-observer variability in the clinic. To address this issue, an automatedmethod utilizing Fully Convolutional Networks (FCN) is proposed to determinemeasurements of fetal head circumference (HC) and biparietal diameter (BPD). AnFCN was trained on approximately 2000 2D ultrasound images of the head withannotations provided by 45 different sonographers during routine screeningexaminations to perform semantic segmentation of the head. An ellipse is fittedto the resulting segmentation contours to mimic the annotation typicallyproduced by a sonographer. The model's performance was compared withinter-observer variability, where two experts manually annotated 100 testimages. Mean absolute model-expert error was slightly better thaninter-observer error for HC (1.99mm vs 2.16mm), and comparable for BPD (0.61mmvs 0.59mm), as well as Dice coefficient (0.980 vs 0.980). Our resultsdemonstrate that the model performs at a level similar to a human expert, andlearns to produce accurate predictions from a large dataset annotated by manysonographers. Additionally, measurements are generated in near real-time at15fps on a GPU, which could speed up clinical workflow for both skilled andtrainee sonographers.
Tarroni G, Oktay O, Sinclair M, et al., 2018, A comprehensive approach for learning-based fully-automated inter-slice motion correction for short-axis cine cardiac MR image stacks, 21st International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI) / 8th Eurographics Workshop on Visual Computing for Biology and Medicine (VCBM), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 268-276, ISSN: 0302-9743
In the clinical routine, short axis (SA) cine cardiac MR (CMR) image stacks are acquired during multiple subsequent breath-holds. If the patient cannot consistently hold the breath at the same position, the acquired image stack will be affected by inter-slice respiratory motion and will not correctly represent the cardiac volume, introducing potential errors in the following analyses and visualisations. We propose an approach to automatically correct inter-slice respiratory motion in SA CMR image stacks. Our approach makes use of probabilistic segmentation maps (PSMs) of the left ventricular (LV) cavity generated with decision forests. PSMs are generated for each slice of the SA stack and rigidly registered in-plane to a target PSM. If long axis (LA) images are available, PSMs are generated for them and combined to create the target PSM; if not, the target PSM is produced from the same stack using a 3D model trained from motion-free stacks. The proposed approach was tested on a dataset of SA stacks acquired from 24 healthy subjects (for which anatomical 3D cardiac images were also available as reference) and compared to two techniques which use LA intensity images and LA segmentations as targets, respectively. The results show the accuracy and robustness of the proposed approach in motion compensation.
Alansary A, Le Folgoc L, Vaillant G, et al., 2018, Automatic view planning with multi-scale deep reinforcement learning agents, International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), Publisher: Springer Verlag, Pages: 277-285, ISSN: 0302-9743
We propose a fully automatic method to find standardizedview planes in 3D image acquisitions. Standard view images are impor-tant in clinical practice as they provide a means to perform biometricmeasurements from similar anatomical regions. These views are often constrained to the native orientation of a 3D image acquisition. Navigating through target anatomy to find the required view plane is tedious and operator-dependent. For this task, we employ a multi-scale reinforcement learning (RL) agent framework and extensively evaluate several DeepQ-Network (DQN) based strategies. RL enables a natural learning paradigm by interaction with the environment, which can be used to mimic experienced operators. We evaluate our results using the distance between the anatomical landmarks and detected planes, and the angles between their normal vector and target. The proposed algorithm is assessed on the mid-sagittal and anterior-posterior commissure planes of brain MRI, and the 4-chamber long-axis plane commonly used in cardiac MRI, achieving accuracy of 1.53mm, 1.98mm and 4.84mm, respectively.
Biffi C, Oktay O, Tarroni G, et al., 2018, Learning interpretable anatomical features through deep generative models: Application to cardiac remodeling, International Conference On Medical Image Computing & Computer Assisted Intervention, Publisher: Springer, Pages: 464-471, ISSN: 0302-9743
Alterations in the geometry and function of the heart define well-established causes of cardiovascular disease. However, current approaches to the diagnosis of cardiovascular diseases often rely on subjective human assessment as well as manual analysis of medical images. Both factors limit the sensitivity in quantifying complex structural and functional phenotypes. Deep learning approaches have recently achieved success for tasks such as classification or segmentation of medical images, but lack interpretability in the feature extraction and decision processes, limiting their value in clinical diagnosis. In this work, we propose a 3D convolutional generative model for automatic classification of images from patients with cardiac diseases associated with structural remodeling. The model leverages interpretable task-specific anatomic patterns learned from 3D segmentations. It further allows to visualise and quantify the learned pathology-specific remodeling patterns in the original input space of the images. This approach yields high accuracy in the categorization of healthy and hypertrophic cardiomyopathy subjects when tested on unseen MR images from our own multi-centre dataset (100%) as well on the ACDC MICCAI 2017 dataset (90%). We believe that the proposed deep learning approach is a promising step towards the development of interpretable classifiers for the medical imaging domain, which may help clinicians to improve diagnostic accuracy and enhance patient risk-stratification.
This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.