692 results found
Ktena SI, Arslan S, Parisot S, et al., 2017, Exploring heritability of functional brain networks with inexact graph matching, IEEE 14th International Symposium on Biomedical Imaging, Publisher: IEEE, Pages: 354-357, ISSN: 1945-7928
Data-driven brain parcellations aim to provide a more accurate representation of an individual's functional connectivity, since they are able to capture individual variability that arises due to development or disease. This renders comparisons between the emerging brain connectivity networks more challenging, since correspondences between their elements are not preserved. Unveiling these correspondences is of major importance to keep track of local functional connectivity changes. We propose a novel method based on graph edit distance for the comparison of brain graphs directly in their domain, that can accurately reflect similarities between individual networks while providing the network element correspondences. This method is validated on a dataset of 116 twin subjects provided by the Human Connectome Project.
Chen L, Bentley P, Rueckert D, 2017, Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks, NeuroImage: Clinical, Vol: 15, Pages: 633-643, ISSN: 2213-1582
Stroke is an acute cerebral vascular disease, which is likely to cause long-term disabilities and death. Acute ischemic lesions occur in most stroke patients. These lesions are treatable under accurate diagnosis and treatments. Although diffusion-weighted MR imaging (DWI) is sensitive to these lesions, localizing and quantifyingthem manually is costly and challenging for clinicians. In this paper, we propose a novel framework to auto-matically segment stroke lesions in DWI. Our framework consists of two convolutional neural networks (CNNs):one is an ensemble of two DeconvNets (Noh et al., 2015), which is the EDD Net; the second CNN is the multi-scale convolutional label evaluation net (MUSCLE Net), which aims to evaluate the lesions detected by the EDD Net in order to remove potential false positives. To the best of our knowledge, it is the first attempt to solve this problem and using both CNNs achieves very good results. Furthermore, we study the network architectures and key configurations in detail to ensure the best performance. It is validated on a large dataset comprising clinical acquired DW images from 741 subjects. A mean accuracy of Dice coefficient obtained is 0.67 in total. The mean Dice scores based on subjects with only small and large lesions are 0.61 and 0.83, respectively. The lesion detection rate achieved is 0.94.
Tong T, Ledig C, Guerrero R, et al., 2017, Five-class differential diagnostics of neurodegenerative diseases using random undersampling boosting, NeuroImage: Clinical, Vol: 15, Pages: 613-624, ISSN: 2213-1582
Differentiating between different types of neurodegenerative diseases is not only crucial in clinical practice when treatment decisions have to be made, but also has a significant potential for the enrichment of clinical trials. The purpose of this study is to develop a classification framework for distinguishing the four most common neurodegenerative diseases, including Alzheimer's disease, frontotemporal lobe degeneration, Dementia with Lewy bodies and vascular dementia, as well as patients with subjective memory complaints. Different biomarkers including features from images (volume features, region-wise grading features) and non-imaging features (CSF measures) were extracted for each subject. In clinical practice, the prevalence of different dementia types is imbalanced, posing challenges for learning an effective classification model. Therefore, we propose the use of the RUSBoost algorithm in order to train classifiers and to handle the class imbalance training problem. Furthermore, a multi-class feature selection method based on sparsity is integrated into the proposed framework to improve the classification performance. It also provides a way for investigating the importance of different features and regions. Using a dataset of 500 subjects, the proposed framework achieved a high accuracy of 75.2% with a balanced accuracy of 69.3% for the five-class classification using ten-fold cross validation, which is significantly better than the results using support vector machine or random forest, demonstrating the feasibility of the proposed framework to support clinical decision making.
Kamnitsas K, Baumgartner C, Ledig C, et al., 2017, Unsupervised domain adaptation in brain lesion segmentation with adversarial networks, Information Processing in Medical Imaging, Publisher: Springer
Significant advances have been made towards building accu-rate automatic segmentation systems for a variety of biomedical applica-tions using machine learning. However, the performance of these systemsoften degrades when they are applied on new data that differ from thetraining data, for example, due to variations in imaging protocols. Man-ually annotating new data for each test domain is not a feasible solution.In this work we investigate unsupervised domain adaptation using ad-versarial neural networks to train a segmentation method which is moreinvariant to differences in the input data, and which does not require anyannotations on the test domain. Specifically, we learn domain-invariantfeatures by learning to counter an adversarial network, which attemptsto classify the domain of the input data by observing the activations ofthe segmentation network. Furthermore, we propose a multi-connecteddomain discriminator for improved adversarial training. Our system isevaluated using two MR databases of subjects with traumatic brain in-juries, acquired using different scanners and imaging protocols. Usingour unsupervised approach, we obtain segmentation accuracies whichare close to the upper bound of supervised domain adaptation.
Tarroni G, Oktay O, Bai W, et al., 2017, Learning-based heart coverage estimation for short-axis cine cardiac MR images, Functional Imaging and Modelling of the Heart (FIMH), Publisher: Springer
The correct acquisition of short axis (SA) cine cardiac MRimage stacks requires the imaging of the full cardiac anatomy betweenthe apex and the mitral valve plane via multiple 2D slices. While in theclinical practice the SA stacks are usually checked qualitatively to en-sure full heart coverage, visual inspection can become infeasible for largeamounts of imaging data that is routinely acquired, e.g. in populationstudies such as the UK Biobank (UKBB). Accordingly, we propose alearning-based technique for the fully-automated estimation of the heartcoverage for SA image stacks. The technique relies on the identificationof cardiac landmarks (i.e. the apex and the mitral valve sides) on twochamber view long axis images and on the comparison of the landmarks’positions to the volume covered by the SA stack. Landmark detection isperformed using a hybrid random forest approach integrating both re-gression and structured classification models. The technique was appliedon 3000 cases from the UKBB and compared to visual assessment. Theobtained results (error rate = 2.3%, sens. = 73%, spec. = 90%) indicatethat the proposed technique is able to correctly detect the vast majorityof the cases with insufficient coverage, suggesting that it could be usedas a fully-automated quality control step for CMR SA image stacks.
Dawes T, Simoes monteiro de marvao A, Shi W, et al., 2017, Machine learning of three-dimensional right ventricular motion enables outcome prediction in pulmonary hypertension: a cardiac MR imaging study, Radiology, Vol: 283, Pages: 381-390, ISSN: 1527-1315
Purpose: To determine if patient survival and mechanisms of right ventricular (RV) failure in pulmonary hypertension (PH) could be predicted using supervised machine learning of three dimensional patterns of systolic cardiac motion. Materials and methods: The study was approved by a research ethics committee and participants gave written informed consent. 256 patients (143 females, mean age 63 ± 17) with newly diagnosed PH underwent cardiac MR imaging, right heart catheterization (RHC) and six minute walk testing (6MWT) with a median follow up of 4.0 years. Semi automated segmentation of short axis cine images was used to create a three dimensional model of right ventricular motion. Supervised principal components analysis identified patterns of systolic motion which were most strongly predictive of survival. Survival prediction was assessed by the difference in median survival time and the area under the curve (AUC) using time dependent receiver operator characteristic for one year survival. Results: At the end of follow up 33% (93/256) died and one underwent lung transplantation. Poor outcome was predicted by a loss of effective contraction in the septum and freewall coupled with reduced basal longitudinal motion. When added to conventional imaging, hemodynamic, functional and clinical markers, three dimensional cardiac motion improved survival prediction (area under the curve 0.73 vs 0.60, p<0.001) and provided greater differentiation by difference in median survival time between high and low risk groups (13.8 vs 10.7 years, p<0.001). Conclusion:Three dimensional motion modeling with machine learning approaches reveal the adaptations in function that occur early in right heart failure and independently predict outcomes in newly diagnosed PH patients.
Toisoul A, Rueckert D, Kainz B, 2017, Accessible GLSL Shader programming, EuroGraphics 2017, ISSN: 1017-4656
Teaching fundamental principles of Computer Graphics requires a thoroughly prepared lecture alongside practical training.Modern graphics programming rarely provides a straightforward application programming interface (API) and the availableAPIs pose high entry barriers to students. Shader-based programming of standard graphics pipelines is often inaccessiblethrough complex setup procedures and convoluted programming environments. In this paper we discuss an undergraduateentry level lecture with its according lab exercises. We present a programming framework that makes interactive graphicsprogramming accessible while allowing to design individual tasks as instructive exercises to solidify the content of individuallecture units. The discussed teaching framework provides a well defined programmable graphics pipeline with geometry shadingstages and image-based post processing functionality based on framebuffer objects. It is open-source and available online.
Valindria V, Lavdas I, Bai W, et al., 2017, Reverse classification accuracy: predicting segmentation performance in the absence of ground truth, IEEE Transactions on Medical Imaging, Vol: 36, Pages: 1597-1606, ISSN: 1558-254X
When integrating computational tools such as au-tomatic segmentation into clinical practice, it is of utmostimportance to be able to assess the level of accuracy on newdata, and in particular, to detect when an automatic methodfails. However, this is difficult to achieve due to absence of groundtruth. Segmentation accuracy on clinical data might be differentfrom what is found through cross-validation because validationdata is often used during incremental method development, whichcan lead to overfitting and unrealistic performance expectations.Before deployment, performance is quantified using differentmetrics, for which the predicted segmentation is compared toa reference segmentation, often obtained manually by an expert.But little is known about the real performance after deploymentwhen a reference is unavailable. In this paper, we introduce theconcept ofreverse classification accuracy(RCA) as a frameworkfor predicting the performance of a segmentation method onnew data. In RCA we take the predicted segmentation froma new image to train a reverse classifier which is evaluatedon a set of reference images with available ground truth. Thehypothesis is that if the predicted segmentation is of good quality,then the reverse classifier will perform well on at least some ofthe reference images. We validate our approach on multi-organsegmentation with different classifiers and segmentation methods.Our results indicate that it is indeed possible to predict the qualityof individual segmentations, in the absence of ground truth. Thus,RCA is ideal for integration into automatic processing pipelines inclinical routine and as part of large-scale image analysis studies.
Arslan S, Ktena SI, Makropoulos A, et al., 2017, Human brain mapping: a systematic comparison of parcellation methods for the human cerebral cortex, Neuroimage, Vol: 170, Pages: 5-30, ISSN: 1095-9572
The macro-connectome elucidates the pathways through which brain regions are structurally connected or functionally coupled to perform a specific cognitive task. It embodies the notion of representing and understanding all connections within the brain as a network, while the subdivision of the brain into interacting functional units is inherent in its architecture. As a result, the definition of network nodes is one of the most critical steps in connectivity network analysis. Although brain atlases obtained from cytoarchitecture or anatomy have long been used for this task, connectivity-driven methods have arisen only recently, aiming to delineate more homogeneous and functionally coherent regions. This study provides a systematic comparison between anatomical, connectivity-driven and random parcellation methods proposed in the thriving field of brain parcellation. Using resting-state functional MRI data from the Human Connectome Project and a plethora of quantitative evaluation techniques investigated in the literature, we evaluate 10 subject-level and 24 groupwise parcellation methods at different resolutions. We assess the accuracy of parcellations from four different aspects: (1) reproducibility across different acquisitions and groups, (2) fidelity to the underlying connectivity data, (3) agreement with fMRI task activation, myelin maps, and cytoarchitectural areas, and (4) network analysis. This extensive evaluation of different parcellations generated at the subject and group level highlights the strengths and shortcomings of the various methods and aims to provide a guideline for the choice of parcellation technique and resolution according to the task at hand. The results obtained in this study suggest that there is no optimal method able to address all the challenges faced in this endeavour simultaneously.
Kamnitsas K, Ferrante E, Parisot S, et al., 2017, DeepMedic for brain tumor segmentation, Second International Workshop, BrainLes 2016, with the Challenges on BRATS, ISLES and mTOP 2016 Held in Conjunction with MICCAI 2016, Publisher: Springer, Pages: 138-149, ISSN: 0302-9743
Accurate automatic algorithms for the segmentation of brain tumours have the potential of improving disease diagnosis, treatment planning, as well as enabling large-scale studies of the pathology. In this work we employ DeepMedic , a 3D CNN architecture previously presented for lesion segmentation, which we further improve by adding residual connections. We also present a series of experiments on the BRATS 2015 training database for evaluating the robustness of the network when less training data are available or less filters are used, aiming to shed some light on requirements for employing such a system. Our method was further benchmarked on the BRATS 2016 Challenge, where it achieved very good performance despite the simplicity of the pipeline.
Karasawa K, Oda M, Kitasaka T, et al., 2017, Multi-atlas pancreas segmentation: Atlas selection based on vessel structure, Medical Image Analysis, Vol: 39, Pages: 18-28, ISSN: 1361-8415
Automated organ segmentation from medical images is an indispensable component for clinical applications such as computer-aided diagnosis (CAD) and computer-assisted surgery (CAS). We utilize a multi-atlas segmentation scheme, which has recently been used in different approaches in the literature to achieve more accurate and robust segmentation of anatomical structures in computed tomography (CT) volume data. Among abdominal organs, the pancreas has large inter-patient variability in its position, size and shape. Moreover, the CT intensity of the pancreas closely resembles adjacent tissues, rendering its segmentation a challenging task. Due to this, conventional intensity-based atlas selection for pancreas segmentation often fails to select atlases that are similar in pancreas position and shape to those of the unlabeled target volume. In this paper, we propose a new atlas selection strategy based on vessel structure around the pancreatic tissue and demonstrate its application to a multi-atlas pancreas segmentation. Our method utilizes vessel structure around the pancreas to select atlases with high pancreatic resemblance to the unlabeled volume. Also, we investigate two types of applications of the vessel structure information to the atlas selection. Our segmentations were evaluated on 150 abdominal contrast-enhanced CT volumes. The experimental results showed that our approach can segment the pancreas with an average Jaccard index of 66.3% and an average Dice overlap coefficient of 78.5%.
Ciller C, De Zanet S, Kamnitsas K, et al., 2017, Multi-channel MRI segmentation of eye structures and tumors using patient-specific features, PLOS ONE, Vol: 12, ISSN: 1932-6203
Retinoblastoma and uveal melanoma are fast spreading eye tumors usually diagnosed by using 2D Fundus Image Photography (Fundus) and 2D Ultrasound (US). Diagnosis and treatment planning of such diseases often require additional complementary imaging to confirm the tumor extend via 3D Magnetic Resonance Imaging (MRI). In this context, having automatic segmentations to estimate the size and the distribution of the pathological tissue would be advantageous towards tumor characterization. Until now, the alternative has been the manual delineation of eye structures, a rather time consuming and error-prone task, to be conducted in multiple MRI sequences simultaneously. This situation, and the lack of tools for accurate eye MRI analysis, reduces the interest in MRI beyond the qualitative evaluation of the optic nerve invasion and the confirmation of recurrent malignancies below calcified tumors. In this manuscript, we propose a new framework for the automatic segmentation of eye structures and ocular tumors in multi-sequence MRI. Our key contribution is the introduction of a pathological eye model from which Eye Patient-Specific Features (EPSF) can be computed. These features combine intensity and shape information of pathological tissue while embedded in healthy structures of the eye. We assess our work on a dataset of pathological patient eyes by computing the Dice Similarity Coefficient (DSC) of the sclera, the cornea, the vitreous humor, the lens and the tumor. In addition, we quantitatively show the superior performance of our pathological eye model as compared to the segmentation obtained by using a healthy model (over 4% DSC) and demonstrate the relevance of our EPSF, which improve the final segmentation regardless of the classifier employed.
Passerat-Palmbach J, Reuillon RR, Leclaire ML, et al., 2017, Reproducible large-scale neuroimaging studies with the OpenMOLE workflow management system, Frontiers in Neuroinformatics, Vol: 11, ISSN: 1662-5196
OpenMOLE is a scientific workflow engine with a strong emphasis on workload distribution.Workflows are designed using a high level Domain Specific Language (DSL) built on top of Scala. It exposes natural parallelism constructs to easily delegate the workload resulting from a workflow to a wide range of distributed computing environments.OpenMOLE hides the complexity of designing complex experiments thanks to its DSL. Users can embed their own applications and scale their pipelines from a small prototype running on their desktop computer to a large-scale study harnessing distributed computing infrastructures, simply by changing a single line in the pipeline definition.The construction of the pipeline itself is decoupled from the execution context. The high-level DSL abstracts the underlying execution environment, contrary to classic shell-script based pipelines. These two aspects allow pipelines to be shared and studies to be replicated across different computing environments.Workflows can be run as traditional batch pipelines or coupled with OpenMOLE's advanced exploration methods in order to study the behaviour of an application, or perform automatic parameter tuning.In this work, we briefly present the strong assets of OpenMOLE and detail recent improvements targeting re-executability of workflows across various Linux platforms. We have tightly coupled OpenMOLE with CARE, a standalone containerisation solution that allows re-executing on a Linux host any application that has been packaged on another Linux host previously.The solution is evaluated against a Python-based pipeline involving packages such as scikit-learn as well as binary dependencies. All were packaged and re-executed successfully on various HPC environments, with identical numerical results (here prediction scores) obtained on each environment. Our results show that the pair formed by OpenMOLE and CARE is a reliable solution to generate reproducible results and re-executable pipelines. A demonstrati
Giannakidis A, Oktay O, Keegan J, et al., Super-resolution Reconstruction of Late Gadolinium Cardiovascular Magnetic Resonance Images using a Residual Convolutional Neural Network, The 25th Scientific Meeting of the International Society for Magnetic Resonance in Medicine (ISMRM 2017)
Schuh A, Makropoulos A, Wright R, et al., A deformable model for the reconstruction of the neonatal cortex, 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI)
We present a method based on deformable meshes for the reconstruction of the cortical surfaces of the developing human brain at the neonatal period. It employs a brain segmentation for the reconstruction of an initial inner cortical surface mesh. Errors in the segmentation resulting from poor tissue contrast in neonatal MRI and partial volume effects are subsequently accounted for by a local edge-based refinement. We show that the obtained surface models define the cortical boundaries more accurately than the segmentation. The surface meshes are further guaranteed to not intersect and subdivide the brain volume into disjoint regions. The proposed method generates topologically correct surfaces which facilitate both a flattening and spherical mapping of the cortex.
Sinclair M, Peressutti D, Puyol-Anton E, et al., 2017, Learning Optimal Spatial Scales for Cardiac Strain Analysis Using a Motion Atlas, 7th International Workshop on Statistical Atlases and Computational Models of the Heart (STACOM), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 57-65, ISSN: 0302-9743
Schlemper J, Caballero J, Hajnal JV, et al., 2017, A Deep Cascade of Convolutional Neural Networks for MR Image Reconstruction, 25th Biennial International Conference on Information Processing in Medical Imaging (IPMI), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 647-658, ISSN: 0302-9743
Baumgartner CF, Oktay O, Rueckert D, 2017, Fully convolutional networks in medical imaging: Applications to image enhancement and recognition, Advances in Computer Vision and Pattern Recognition, Pages: 159-179
© Springer International Publishing Switzerland 2017. Convolutional neural networks (CNNs) are hierarchical models that have immense representational capacity and have been successfully applied to computer vision problems including object localisation, classification and super-resolution. A particular example of CN Nmodels, knownas fully convolutional network (FCN), has been shown to offer improved computational efficiency and representation learning capabilities due to simplermodel parametrisation and spatial consistency of extracted features. In this chapter, we demonstrate the power and applicability of this particular model on two medical imaging tasks, image enhancement via super-resolution and image recognition. In both examples, experimental results show that FCN models can significantly outperform traditional learning-based approaches while achieving real-time performance. Additionally, we demonstrate that the proposed image classification FCN model can be used in organ localisation task as well without requiring additional training data.
Oda M, Shimizu N, Roth HR, et al., 2017, 3D FCN Feature Driven Regression Forest-Based Pancreas Localization and Segmentation, 3rd MICCAI International Workshop on Deep Learning in Medical Image Analysis (DLMIA) / 7th International Workshop on Multimodal Learning for Clinical Decision Support (ML-CDS), Publisher: SPRINGER INTERNATIONAL PUBLISHING AG, Pages: 222-230, ISSN: 0302-9743
Ktena SI, Parisot S, Ferrante E, et al., 2017, Distance Metric Learning Using Graph Convolutional Networks: Application to Functional Brain Networks., Publisher: Springer, Pages: 469-477
Lorch B, Vaillant G, Baumgartner C, et al., 2017, Automated Detection of Motion Artefacts in MR Imaging Using Decision Forests., J Med Eng, Vol: 2017, ISSN: 2314-5137
The acquisition of a Magnetic Resonance (MR) scan usually takes longer than subjects can remain still. Movement of the subject such as bulk patient motion or respiratory motion degrades the image quality and its diagnostic value by producing image artefacts like ghosting, blurring, and smearing. This work focuses on the effect of motion on the reconstructed slices and the detection of motion artefacts in the reconstruction by using a supervised learning approach based on random decision forests. Both the effects of bulk patient motion occurring at various time points in the acquisition on head scans and the effects of respiratory motion on cardiac scans are studied. Evaluation is performed on synthetic images where motion artefacts have been introduced by altering the k-space data according to a motion trajectory, using the three common k-space sampling patterns: Cartesian, radial, and spiral. The results suggest that a machine learning approach is well capable of learning the characteristics of motion artefacts and subsequently detecting motion artefacts with a confidence that depends on the sampling pattern.
Robinson E, Glocker B, Rajchl M, et al., 2016, Discrete Optimisation for Group-wise Cortical Surface Atlasing, International Workshop on Biomedical Image Registration, Publisher: IEEE, ISSN: 2160-7516
This paper presents a novel method for cortical surfaceatlasing. Group-wise registration is performed through adiscrete optimisation framework that seeks to simultaneouslyimprove pairwise correspondences between surfacefeature sets, whilst minimising a global cost relating to therank of the feature matrix. It is assumed that when fullyaligned, features will be highly linearly correlated, andthus have low rank. The framework is regularised throughuse of multi-resolution control point grids and higher-ordersmoothness terms, calculated by considering deformationstrain for displacements of triplets of points. Accordinglythe discrete framework is solved through high-order cliquereduction. The framework is tested on cortical foldingbased alignment, using data from the Human ConnectomeProject. Preliminary results indicate that group-wise alignmentimproves folding correspondences, relative to registrationbetween all pair-wise combinations, and registrationto a global average template.
Gomez A, Oktay O, Rueckert D, et al., 2016, Regional differences in end-diastolic volumes between 3D echo and CMR in HLHS patients, Frontiers in Pediatrics, Vol: 4, ISSN: 2296-2360
Ultrasound is commonly thought to underestimate ventricular volumes compared to magnetic resonance imaging (MRI), although the reason for this and the spatial distribution of the volume difference is not well understood. In this paper, we use landmark-based image registration to spatially align MRI and ultrasound images from patients with hypoplastic left heart syndrome and carry out a qualitative and quantitative spatial comparison of manual segmentations of the ventricular volume obtained from the respective modalities. In our experiments, we have found a trend showing volumes estimated from ultrasound to be smaller than those obtained from MRI (by approximately up to 20 ml), and that important contributors to this difference are the presence of artifacts such as shadows in the echo images and the different criteria to include or exclude image features as part of the ventricular volume.
Xie Z, 2016, Machine learning for efficient recognition of anatomical structures and abnormalities in biomedical images
Schafer S, de Marvao A, Adami E, et al., 2016, Titin truncating variants affect heart function in disease cohorts and the general population, Nature Genetics, Vol: 49, Pages: 46-53, ISSN: 1546-1718
Titin-truncating variants (TTNtv) commonly cause dilated cardiomyopathy (DCM). TTNtv are also encountered in ~1% of the general population, where they may be silent, perhaps reflecting allelic factors. To better understand TTNtv, we integrated TTN allelic series, cardiac imaging and genomic data in humans and studied rat models with disparate TTNtv. In patients with DCM, TTNtv throughout titin were significantly associated with DCM. Ribosomal profiling in rat showed the translational footprint of premature stop codons in Ttn, TTNtv-position-independent nonsense-mediated degradation of the mutant allele and a signature of perturbed cardiac metabolism. Heart physiology in rats with TTNtv was unremarkable at baseline but became impaired during cardiac stress. In healthy humans, machine-learning-based analysis of high-resolution cardiac imaging showed TTNtv to be associated with eccentric cardiac remodeling. These data show that TTNtv have molecular and physiological effects on the heart across species, with a continuum of expressivity in health and disease.
Rajchl M, Lee MCH, Oktay O, et al., 2016, DeepCut: object segmentation from bounding box annotations using convolutional neural networks, IEEE Transactions on Medical Imaging, Vol: 36, Pages: 674-683, ISSN: 0278-0062
In this paper, we propose DeepCut, a method to obtain pixelwise object segmentations given an image dataset labelled weak annotations, in our case bounding boxes. It extends the approach of the well-known GrabCut method to include machine learning by training a neural network classifier from bounding box annotations. We formulate the problem as an energy minimisation problem over a densely-connected conditional random field and iteratively update the training targets to obtain pixelwise object segmentations. Additionally, we propose variants of the DeepCut method and compare those to a naïve approach to CNN training under weak supervision. We test its applicability to solve brain and lung segmentation problems on a challenging fetal magnetic resonance dataset and obtain encouraging results in terms of accuracy.
Kamnitsas K, Ledig C, Newcombe VFJ, et al., 2016, Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation, Medical Image Analysis, Vol: 36, Pages: 61-78, ISSN: 1361-8423
We propose a dual pathway, 11-layers deep, three-dimensional ConvolutionalNeural Network for the challenging task of brain lesion segmentation. Thedevised architecture is the result of an in-depth analysis of the limitations ofcurrent networks proposed for similar applications. To overcome the computationalburden of processing 3D medical scans, we have devised an efficientand effective dense training scheme which joins the processing of adjacentimage patches into one pass through the network while automatically adaptingto the inherent class imbalance present in the data. Further, we analyzethe development of deeper, thus more discriminative 3D CNNs. In order toincorporate both local and larger contextual information, we employ a dualpathway architecture that processes the input images at multiple scales simultaneously.For post-processing of the network’s soft segmentation, we use a3D fully connected Conditional Random Field which effectively removes falsepositives. Our pipeline is extensively evaluated on three challenging tasks oflesion segmentation in multi-channel MRI patient data with traumatic braininjuries, brain tumors, and ischemic stroke. We improve on the state-of-theartfor all three applications, with top ranking performance on the publicbenchmarks BRATS 2015 and ISLES 2015. Our method is computationallyefficient, which allows its adoption in a variety of research and clinicalsettings. The source code of our implementation is made publicly available
Giannakidis A, Kamnitsas K, Spadotto V, et al., 2017, Fast fully automatic segmentation of the severely abnormal human right ventricle from cardiovascular magnetic resonance images using a multi-scale 3D convolutional neural network, The 12th IEEE International Conference on Signal Image Technology & Internet Systems (IEEE SITIS 2016), Publisher: IEEE
Cardiac magnetic resonance (CMR) is regardedas the reference examination for cardiac morphology intetralogy of Fallot (ToF) patients allowing images of high spa-tial resolution and high contrast. The detailed knowledge ofthe right ventricular anatomy is critical in ToF management.The segmentation of the right ventricle (RV) in CMR imagesfrom ToF patients is a challenging task due to the high shapeand image quality variability. In this paper we propose a fullyautomatic deep learning-based framework to segment the RVfrom CMR anatomical images of the whole heart. We adopt a3D multi-scale deep convolutional neural network to identifypixels that belong to the RV. Our robust segmentationframework was tested on 26 ToF patients achieving a Dicesimilarity coefficient of 0.8281±0.1010 with reference tomanual annotations performed by expert cardiologists. Theproposed technique is also computationally efficient, whichmay further facilitate its adoption in the clinical routine.
Ktena SI, Parisot S, Passerat-Palmbach J, et al., 2016, Comparison of brain networks with unknown correspondences, MICCAI Workshop on Brain Analysis using COnnectivity Networks (BACON) 2016, Publisher: BACON 2016
Graph theory has drawn a lot of attention in the field of Neuroscience duringthe last decade, mainly due to the abundance of tools that it provides toexplore the interactions of elements in a complex network like the brain. Thelocal and global organization of a brain network can shed light on mechanismsof complex cognitive functions, while disruptions within the network can belinked to neurodevelopmental disorders. In this effort, the construction of arepresentative brain network for each individual is critical for furtheranalysis. Additionally, graph comparison is an essential step for inference andclassification analyses on brain graphs. In this work we explore a method basedon graph edit distance for evaluating graph similarity, when correspondencesbetween network elements are unknown due to different underlying subdivisionsof the brain. We test this method on 30 unrelated subjects as well as 40 twinpairs and show that this method can accurately reflect the higher similaritybetween two related networks compared to unrelated ones, while identifying nodecorrespondences.
Peressutti D, Sinclair M, Bai W, et al., 2016, A framework for combining a motion atlas with non-motion information to learn clinically useful biomarkers: Application to cardiac resynchronisation therapy response prediction, MEDICAL IMAGE ANALYSIS, Vol: 35, Pages: 669-684, ISSN: 1361-8415
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.