Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

Diffusion Spectrum Imaging (DSI) offers detailed information on complex distributions of intravoxel fiber orientations at the expense of extremely long imaging times (~1 hour). It is possible to accelerate DSI by sub-Nyquist sampling of the

q

-space followed by nonlinear reconstruction to estimate the diffusion probability density functions (pdfs). Recent work by Menzel et al. imposed sparsity constraints on the pdfs under wavelet and Total Variation (TV) transforms. As the performance of Compressed Sensing (CS) reconstruction depends strongly on the level of sparsity in the selected transform space, a dictionary specifically tailored for sparse representation of diffusion pdfs can yield higher fidelity results. To our knowledge, this work is the first application of adaptive dictionaries in DSI, whereby we reduce the scan time of whole brain DSI acquisition from 50 to 17 min while retaining high image quality.

In vivo

experiments were conducted with the novel 3T

Connectome

MRI, whose strong gradients are particularly suited for DSI. The RMSE from the proposed reconstruction is up to 2 times lower than that of Menzel et al.’s method, and is actually comparable to that of the fully-sampled 50 minute scan. Further, we demonstrate that a dictionary trained using pdfs from a single slice of a particular subject generalizes well to other slices from the same subject, as well as to slices from another subject.

In this work, we propose an original and efficient approach to exploit the ability of Compressed Sensing (CS) to recover Diffusion MRI (dMRI) signals from a limited number of samples while efficiently recovering important diffusion features such as the Ensemble Average Propagator (EAP) and the Orientation Distribution Function (ODF). Some attempts to sparsely represent the diffusion signal have already been performed. However and contrarly to what has been presented in CS dMRI, in this work we propose and advocate the use of a well adapted learned dictionary and show that it leads to a sparser signal estimation as well as to an efficient reconstruction of very important diffusion features. We first propose to learn and design a sparse and parametric dictionary from a set of training diffusion data. Then, we propose a framework to analytically estimate in closed form two important diffusion features : the EAP and the ODF. Various experiments on synthetic, phantom and human brain data have been carried out and promising results with reduced number of atoms have been obtained on diffusion signal reconstruction, thus illustrating the added value of our method over state-of-the-art SHORE and SPF based approaches.

Diffusion-weighted imaging (DWI), while giving rich information about brain circuitry, is often limited by insufficient spatial resolution and low signal-to-noise ratio (SNR). This paper describes an algorithm that will increase the resolution of DW images beyond the scan resolution, allowing for a closer investigation of fiber structures and more accurate assessment of brain connectivity. The algorithm is capable of generating a dense vector-valued field, consisting of diffusion data associated with the full set of diffusion-sensitizing gradients. The fundamental premise is that, to best preserve information, interpolation should always be performed along fiber streamlines. To achieve this, at each spatial location, we probe neighboring voxels in various directions to gather diffusion information for data reconstruction. Based on the fiber orientation distribution (FOD), directions that are more likely to be traversed by fibers will be given greater weights during interpolation and vice versa. This ensures that data reconstruction is only contributed by diffusion data coming from fibers that are aligned with a specific direction. This approach respects local fiber structures and prevents blurring resulting from averaging of data from significantly misaligned fibers. Evaluations suggest that this algorithm yields results with significantly less blocking artifacts, greater smoothness in anatomical structures, and markedly improved structural visibility.

A new method for the geometrical averaging of labels or landmarks is presented. This method expands the shape-based averaging [1] framework from an Euclidean to a geodesic based distance, incorporating a spatially varying similarity term as time cost. This framework has unique geometrical properties, making it ideal for propagating very small structures following rigorous labelling protocols. The method is used to automate the seeding and way-pointing of optic radiation tractography in DTI imaging. The propagated seeds and waypoints follow a strict clinical protocol by being geometrically constrained to one single slice and by guaranteeing spatial contiguity. The proposed method not only reduces the fragmentation of the propagated areas but also significantly increases the seed positioning accuracy and subsequent tractography results when compared to state-of-the-art label fusion techniques.

The geometry of white matter tracts is of increased interest for a variety of neuroscientific investigations, as it is a feature reflective of normal neurodevelopment and disease factors that may affect it. In this paper, we introduce a novel method for computing multi-scale fibre tract shape and geometry based on the differential geometry of curve sets. By measuring the variation of a curve’s tangent vector at a given point in all directions orthogonal to the curve, we obtain a 2D “dispersion distribution function” at that point. That is, we compute a function on the unit circle which describes fibre dispersion, or fanning, along each direction on the circle. Our formulation is then easily incorporated into a continuous scale-space framework. We illustrate our method on different fibre tracts and apply it to a population study on hemispheric lateralization in healthy controls. We conclude with directions for future work.

Acquisition geometries for tomographic reconstruction are usually densely sampled in order to keep the underlying linear system used in iterative reconstruction as well–posed as possible. While this objective is easily enforced in imaging systems with gantries, this issue is more critical for intra–operative setups using freehand–guided data sensing. This paper investigates an incremental method to monitor the numerical condition of the system based on the singular value decomposition of the system matrix, and presents an approach to find optimal detector positions via a randomized optimization scheme. The feasibility of this approach is demonstrated using simulations of an intra–operative functional imaging setup and actual robot–controlled phantom experiments.

Many successful segmentation algorithms are based on Bayesian models in which prior anatomical knowledge is combined with the available image information. However, these methods typically have many free parameters that are estimated to obtain point estimates only, whereas a faithful Bayesian analysis would also consider all possible alternate values these parameters may take. In this paper, we propose to incorporate the uncertainty of the free parameters in Bayesian segmentation models more accurately by using Monte Carlo sampling. We demonstrate our technique by sampling atlas warps in a recent method for hippocampal subfield segmentation, and show a significant improvement in an Alzheimer’s disease classification task. As an additional benefit, the method also yields informative “error bars” on the segmentation results for each of the individual sub-structures.

The spatio-temporal coherence in data plays an important role in echocardiographic segmentation. While learning offline dynamical priors from databases has received considerable attention, these priors may not be suitable for post-infarct patients and children with congenital heart disease. This paper presents a dynamical appearance model (DAM) driven by individual inherent data coherence. It employs multiscale sparse representation of local appearance, learns online multiscale appearance dictionaries as the image sequence is segmented sequentially, and integrates a spectrum of complementary multiscale appearance information including intensity, multiscale local appearance, and dynamical shape predictions. It overcomes the limitations of database-driven statistical models and applies to a broader range of subjects. Results on 26 4D canine echocardiographic images acquired from both healthy and post-infarct subjects show that our method significantly improves segmentation accuracy and robustness compared to a conventional intensity model and our previous single-scale sparse representation method.

Kidney segmentation in 3D CT images allows extracting useful information for nephrologists. For practical use in clinical routine, such an algorithm should be fast, automatic and robust to contrast-agent enhancement and fields of view. By combining and refining state-of-the-art techniques (random forests and template deformation), we demonstrate the possibility of building an algorithm that meets these requirements. Kidneys are localized with random forests following a coarse-to-fine strategy. Their initial positions detected with global contextual information are refined with a cascade of local regression forests. A classification forest is then used to obtain a probabilistic segmentation of both kidneys. The final segmentation is performed with an implicit template deformation algorithm driven by these kidney probability maps. Our method has been validated on a highly heterogeneous database of 233 CT scans from 89 patients. 80 % of the kidneys were accurately detected and segmented (Dice coefficient > 0.90) in a few seconds per volume.

Methods that leverage neighbourhood structures in high-dimensional image spaces have recently attracted attention. These approaches extract information from a new image using its “neighbours” in the image space equipped with an application-specific distance. Finding the neighbourhood of a given image is challenging due to large dataset sizes and costly distance evaluations. Furthermore, automatic neighbourhood search for a new image is currently not possible when the distance is based on ground truth annotations. In this article we present a general and efficient solution to these problems. “Neighbourhood Approximation Forests” (NAF) is a supervised learning algorithm that approximates the neighbourhood structure resulting from an arbitrary distance. As NAF uses only image intensities to infer neighbours it can also be applied to distances based on ground truth annotations. We demonstrate NAF in two scenarios: i) choosing neighbours with respect to a deformation-based distance, and ii) age prediction from brain MRI. The experiments show NAF’s approximation quality, computational advantages and use in different contexts.

A novel approach to the problem of locating and recognizing anatomical structures of interest in ultrasound (US) video is proposed. While addressing this challenge may be beneficial to US examinations in general, it is particularly useful in situations where portable US probes are used by less experienced personnel. The proposed solution is based on the hypothesis that, rather than their appearance in a single image, anatomical structures are most distinctively characterized by the variation of their appearance as the transducer moves. By drawing on recent advances in the non-linear modeling of video appearance and motion, using an extension of dynamic textures, successful location and recognition is demonstrated on two phantoms. We further analyze computational demands and preliminarily explore insensitivity to anatomic variations.

Extending mutual information (MI), which has been widely used as a similarity measure for rigid registration of multi-modal images, to deformable registration is an active field of research. We propose a self-similarity weighted graph-based implementation of

α

-mutual information (

α

-MI) for nonrigid image registration. The new

Se

lf

S

imilarity

$\underline \alpha$

-

MI

(SeSaMI) metric takes local structures into account and is robust against signal non-stationarity and intensity distortions. We have used SeSaMI as the similarity measure in a regularized cost function with B-spline deformation field. Since the gradient of SeSaMI can be derived analytically, the cost function can be efficiently optimized using stochastic gradient descent. We show that SeSaMI produces a robust and smooth cost function and outperforms the state of the art statistical based similarity metrics in simulation and using data from image-guided neurosurgery.

This paper presents a new approach for the robust alignment and interpretation of 3D anatomical structures with large and localized shape differences. In such situations, existing techniques based on the well-known Procrustes analysis can be significantly affected due to the introduced non-Gaussian distribution of the residuals. In the proposed technique, influential points that induce large dissimilarities are identified and displaced with the aim to obtain an intermediate template with an improved distribution of the residuals. The key element of the algorithm is the use of pose invariant shape variables to robustly guide both the influential point detection and displacement steps. The intermediate template is then used as the basis for the estimation of the final pose parameters between the source and destination shapes, enabling to effectively highlight the regional differences of interest. The validation using synthetic and real datasets of different morphologies demonstrates robustness up-to 50% regional differences and potential for shape classification.

Transrectal ultrasound (TRUS) facilitates intra-treatment delineation of the prostate gland (PG) to guide insertion of brachytherapy seeds, but the prostate substructure and apex are not always visible which may make the seed placement sub-optimal. Based on an elastic model of the prostate created from MRI, where the prostate substructure and apex are clearly visible, we use a Bayesian approach to estimate the posterior distribution on deformations that aligns the pre-treatment MRI with intra-treatment TRUS. Without apex information in TRUS, the posterior prediction of the location of the prostate boundary, and the prostate apex boundary in particular, is mainly determined by the pseudo stiffness hyper-parameter of the prior distribution. We estimate the optimal value of the stiffness through likelihood maximization that is sensitive to the accuracy as well as the precision of the posterior prediction at the apex boundary. From a data-set of 10 pre- and intra-treatment prostate images with ground truth delineation of the total PG, 4 cases were used to establish an optimal stiffness hyper-parameter when 15% of the prostate delineation was removed to simulate lack of apex information in TRUS, while the remaining 6 cases were used to cross-validate the registration accuracy and uncertainty over the PG and in the apex.

Deformable image registration poses a highly non-convex optimisation problem. Conventionally, medical image registration techniques rely on continuous optimisation, which is prone to local minima. Recent advances in the mathematics and new programming methods enable these disadvantages to be overcome using discrete optimisation. In this paper, we present a new technique

deeds

, which employs a discrete

de

ns

e

d

isplacement

s

ampling for the deformable registration of high resolution CT volumes. The image grid is represented as a minimum spanning tree. Given these constraints a global optimum of the cost function can be found efficiently using dynamic programming, which enforces the smoothness of the deformations. Experimental results demonstrate the advantages of

deeds

: the registration error for the challenging registration of inhale and exhale pulmonary CT scans is significantly lower than for two state-of-the-art registration techniques, especially in the presence of large deformations and sliding motion at lung surfaces.

We present what we believe to be the first investigation into unbiased multi-subject registration of whole brain diffusion tractography of the white matter. To our knowledge, this is also the first entropy-based objective function applied to fiber tract registration. To define the probability of fiber trajectories for the computation of entropy, we take advantage of a pairwise fiber distance used as the basis for a Gaussian-like kernel. By employing several values of the kernel’s scale parameter, the method is inherently multi-scale. Results of experiments using synthetic and real datasets demonstrate the potential of the method for simultaneous joint registration of tractography.

We propose a method for deformable registration based on learning the manifolds of individual brain regions. Recent publications on registration of medical images advocate the use of manifold learning in order to confine the search space to anatomically plausible deformations. Existing methods construct manifolds based on a single metric over the entire image domain thus frequently miss regional brain variations. We address this issue by first learning manifolds for specific regions and then computing region-specific deformations from these manifolds. We then determine deformations for the entire image domain by learning the global manifold in such a way that it preserves the region-specific deformations. We evaluate the accuracy of our method by applying it to the LPBA40 dataset and measuring the overlap of the deformed segmentations. The result shows significant improvement in registration accuracy on cortex regions compared to other state of the art methods.

Fiducial-based registration is often utilized in image guided surgery because of its simplicity and speed. The assessment of target registration error when using this technique, however, is difficult. Although the distribution of the target registration error can be estimated given the fiducial configuration and an estimation of the fiducial localization error, the target registration error for a specific registration is uncorrelated with the fiducial registration error. Fiducial registration error is thus an unreliable predictor of the target registration error for a particular case. In this work, we present a new method to estimate the quality of a fiducial-based registration and show that our measure is correlated to the target registration error and that it can be used to reduce registration error caused by fiducial localization error. This has direct implication on the attainable accuracy of fiducial-based registration methods.

We present a fast and robust supervised algorithm for labeling anatomical airway trees, based on geodesic distances in a geometric tree-space. Possible branch label configurations for a given tree are evaluated based on distances to a training set of labeled trees. In tree-space, the tree topology and geometry change continuously, giving a natural way to automatically handle anatomical differences and noise. The algorithm is made efficient using a hierarchical approach, in which labels are assigned from the top down. We only use features of the airway centerline tree, which are relatively unaffected by pathology.

A thorough leave-one-patient-out evaluation of the algorithm is made on 40 segmented airway trees from 20 subjects labeled by 2 medical experts. We evaluate accuracy, reproducibility and robustness in patients with Chronic Obstructive Pulmonary Disease (COPD). Performance is statistically similar to the inter- and intra-expert agreement, and we found no significant correlation between COPD stage and labeling accuracy.

Groupwise non-rigid registration is an important technique in medical image analysis. Recent studies show that its accuracy can be greatly improved by explicitly providing good initialisation. This is achieved by seeking a sparse correspondence using a parts+geometry model. In this paper we show that a single parts+geometry model is unlikely to establish consistent sparse correspondence for complex objects, and that better initialisation can be achieved using a set of models. We describe how to combine the strengths of multiple models, and demonstrate that the method gives state-of-the-art performance on three datasets, with the most significant improvement on the most challenging.

In this paper a novel groupwise registration algorithm is proposed for the unbiased registration of a

large

number of densely sampled point clouds. The method fits an evolving mean shape to each of the example point clouds thereby minimizing the total deformation. The registration algorithm alternates between a computationally expensive, but parallelizable, deformation step of the mean shape to each example shape and a very inexpensive step updating the mean shape.

The algorithm is evaluated by comparing it to a state of the art registration algorithm [5]. Bone surfaces of wrists, segmented from CT data with a voxel size of 0.3×0.3×0.3 mm

3

, serve as an example test set. The negligible bias and registration error of about 0.12 mm for the proposed algorithm are similar to those in [5]. However, current

point cloud

registration algorithms usually have computational and memory costs that increase quadratically with the number of point clouds, whereas the proposed algorithm has linearly increasing costs, allowing the registration of a much larger number of shapes: 48 versus 8, on the hardware used.

In this paper we propose an auditory stimulation and Near Infra-Red Spectroscopy (NIRS) hemodynamic changes acquisition protocol for preterm neonates. This study is designed to assess the specific characteristics of neurovascular coupling to auditory stimuli in healthy and ill neonate brains. The method could lead to clinical application in Intra-Ventricular Hemorrhage (IVH) diagnosis along with other techniques such as EEG. We propose a realistic head model creation with all useful head structures and brain tissues including the neonate fontanel for more accurate results from NIRS signals modeling. We also design a 3D imaging tool for dynamic mapping and analysis of brain activation onto the cortex surface. Results show significant differences in oxy-hemoglobin between healthy neonates and subjects with IVH.

Identifying brain hemodynamics in event-related functional MRI (fMRI) data is a crucial issue to disentangle the vascular response from the neuronal activity in the BOLD signal. This question is usually addressed by estimating the so-called Hemodynamic Response Function (HRF). Voxelwise or region-/parcelwise inference schemes have been proposed to achieve this goal but so far all known contributions commit to pre-specified spatial supports for the hemodynamic territories by defining these supports either as individual voxels or a priori fixed brain parcels. In this paper, we introduce a Joint Parcellation-Detection-Estimation (JPDE) procedure that incorporates an adaptive parcel identification step based upon local hemodynamic properties. Efficient inference of both evoked activity, HRF shapes and

supports

is then achieved using variational approximations. Validation on synthetic and real fMRI data demonstrate the JPDE performance over standard detection estimation schemes and suggest it as a new brain exploration tool.

Identifying functional networks from resting-state functional MRI is a challenging task, especially for multiple subjects. Most current studies estimate the networks in a sequential approach, i.e., they identify each individual subject’s network independently to other subjects, and then estimate the group network from the subjects networks. This one-way flow of information prevents one subject’s network estimation benefiting from other subjects. We propose a hierarchical Markov Random Field model, which takes into account both the within-subject spatial coherence and between-subject consistency of the network label map. Both population and subject network maps are estimated simultaneously using a Gibbs sampling approach in a Monte Carlo Expectation Maximization framework. We compare our approach to two alternative groupwise fMRI clustering methods, based on K-means and Normalized Cuts, using both synthetic and real fMRI data. We show that our method is able to estimate more consistent subject label maps, as well as a stable group label map.

We propose a

metamorphic

geodesic regression approach approximating spatial transformations for image time-series while simultaneously accounting for intensity changes. Such changes occur for example in magnetic resonance imaging (MRI) studies of the developing brain due to myelination. To simplify computations we propose an approximate metamorphic geodesic regression formulation that only requires pairwise computations of image metamorphoses. The approximated solution is an appropriately weighted average of initial momenta. To obtain initial momenta reliably, we develop a shooting method for image metamorphosis.

We contribute a novel and interpretable dimensionality reduction strategy,

eigenanatomy

, that is tuned for neuroimaging data. The method approximates the eigendecomposition of an image set with basis functions (the eigenanatomy vectors) that are

sparse

,

unsigned

and are

anatomically clustered

. We employ the eigenanatomy vectors as anatomical predictors to improve detection power in morphometry. Standard voxel-based morphometry (VBM) analyzes imaging data voxel-by-voxel—and follows this with cluster-based or voxel-wise multiple comparisons correction methods to determine significance. Eigenanatomy reverses the standard order of operations by first clustering the voxel data and then using standard linear regression in this reduced dimensionality space. As with traditional region-of-interest (ROI) analysis, this strategy can greatly improve detection power. Our results show that eigenanatomy provides a principled objective function that leads to localized, data-driven regions of interest. These regions improve our ability to quantify biologically plausible rates of cortical change in two distinct forms of neurodegeneration. We detail the algorithm and show experimental evidence of its efficacy.

Accurate localization of functionally meaningful Regions of Interests (ROIs) from fMRI data is critically important to functional brain imaging. A variety of established approaches such as General Linear Model (GLM) have been widely used in the community. How to determine the optimal location and size of an fMRI-derived ROI, however, remains an open, challenging problem. This paper presents a novel individualized optimization algorithm that simultaneously optimizes the locations and sizes of fMRI-derived ROIs by maximizing the coherences of their functional interaction patterns with respect to the block-based paradigm. As an alternative ROI optimization approach using functional interaction patterns, the algorithm was applied on a working memory task-based fMRI dataset and the experimental results are promising.

Statistical analysis of shapes, performed by constructing an atlas composed of an average model of shapes within a population and associated deformation maps, is a fundamental aspect of medical imaging studies. Usual methods for constructing a shape atlas require point correspondences across subjects, which are difficult in practice. By contrast, methods based on currents do not require correspondence. However, existing atlas construction methods using currents suffer from two limitations. First, the template current is not in the form of a topologically correct mesh, which makes direct analysis on shapes difficult. Second, the deformations are parametrized by vectors at the same location as the normals of the template current which often provides a parametrization that is more dense than required. In this paper, we propose a novel method for constructing shape atlases using currents where topology of the template is preserved and deformation parameters are optimized independently of the shape parameters. We use an

L

1

-type prior that enables us to adaptively compute sparse and low dimensional parameterization of deformations. We show an application of our method for comparing anatomical shapes of patients with Down’s syndrome and healthy controls, where the sparse parametrization of diffeomorphisms decreases the parameter dimension by one order of magnitude.

Connectivity matrices obtained from various modalities (DTI, MEG and fMRI) provide a unique insight into brain processes. Their high dimensionality necessitates the development of methods for population-based statistics, in the face of small sample sizes. In this paper, we present such a method applicable to functional connectivity networks, based on identifying the basis of dominant connectivity components that characterize the patterns of brain pathology and population variation. Projection of individual connectivity matrices into this basis allows for dimensionality reduction, facilitating subsequent statistical analysis. We find dominant components for a collection of connectivity matrices by using the projective non-negative component analysis technique which ensures that the components have non-negative elements and are non-negatively combined to obtain individual subject networks, facilitating interpretation. We demonstrate the feasibility of our novel framework by applying it to simulated connectivity matrices as well as to a clinical study using connectivity matrices derived from resting state magnetoencephalography (MEG) data in a population of subjects diagnosed with autism spectrum disorder (ASD).

Neuroimage analysis based on machine learning technologies has been widely employed to assist the diagnosis of brain diseases such as Alzheimer’s disease and its prodromal stage - mild cognitive impairment. One of the major problems in brain image analysis involves learning the most relevant features from a huge set of raw imaging features, which are far more numerous than the training samples. This makes the tasks of both disease classification and interpretation extremely challenging. Sparsecoding via L1 regularization, such as Lasso, can provide an effective way to select the most relevant features for alleviating the curse of dimensionality and achieving more accurate classification. However, the selected features may distribute randomly throughout the whole brain, although in reality disease-induced abnormal changes often happen in a few contiguous regions. To address this issue, we investigate a tree-guided sparse coding method to identify grouped imaging features in the brain regions for guiding disease classification and interpretation. Spatial relationships of the image structures are imposed during sparse coding with a tree-guided regularization. Our experimental results on the ADNI dataset show that the tree-guided sparse coding method not only achieves better classification accuracy, but also allows for more meaningful diagnosis of brain diseases compared with the conventional L1-regularized Lasso.

Typical cohorts in brain imaging studies are not large enough for systematic testing of all the information contained in the images. To build testable working hypotheses, investigators thus rely on analysis of previous work, sometimes formalized in a so-called

meta-analysis

. In brain imaging, this approach underlies the specification of regions of interest (ROIs) that are usually selected on the basis of the coordinates of previously detected effects. In this paper, we propose to use a database of images, rather than coordinates, and frame the problem as

transfer learning

: learning a discriminant model on a reference task to apply it to a different but related new task. To facilitate statistical analysis of small cohorts, we use a sparse discriminant model that selects predictive voxels on the reference task and thus provides a principled procedure to define ROIs. The benefits of our approach are twofold. First it uses the reference database for prediction,

i.e.

to provide potential biomarkers in a clinical setting. Second it increases statistical power on the new task. We demonstrate on a set of 18 pairs of functional MRI experimental conditions that our approach gives good prediction. In addition, on a specific transfer situation involving different scanners at different locations, we show that voxel selection based on transfer learning leads to higher detection power on small cohorts.

Preterm birth increases the risk of perinatal brain injury and is believed to initiate a cascade of processes causing white matter damage resulting in subsequent neurological deficit; neonatal magnetic resonance imaging provides a number of potential biomarkers of this deficit. In this work we unify measures of the cortical folding pattern and of white matter integrity to establish correlation between grey and white matter derived properties. Diffusion weighted MRI has revealed that the cortical grey matter in the extremely preterm period exhibits a strong transient radial organisation suggesting neuronal axons are orientated towards the underlying white matter. This effect is lost during cortical maturation and is considered no longer visible on MRI at term equivalent age. Here we show that, in a group of 19 infants, radial organisation in the cortical grey matter remains detectable at term-equivalent age and that there is a strong anterior-posterior asymmetry. A group of three infants with moderate or severe abnormal white matter abnormality have significantly higher cortical grey matter radial organisation (

p

 < 0.02), higher grey matter FA (

p

 < 0.01) and a lower measure of cortical complexity (

p

 < 0.03) than infants with normal or mild abnormal white matter abnormality; all measures associated with the preterm phenotype before term equivalent age. The novel combination of state-of-the-art imaging techniques, analysing grey-matter based spatial characteristics, may provide insight into the mechanism of neurodevelopmental deficits seen in infants with abnormal MR imaging at term equivalent age.

Sparse learning has recently received increasing attentions in neuroimaging research such as brain disease diagnosis and progression. Most existing studies focus on cross-sectional analysis, i.e., learning a sparse model based on single time-point of data. However, in some brain imaging applications, multiple time-points of data are often available, thus longitudinal analysis can be performed to better uncover the underlying disease progression patterns. In this paper, we propose a novel temporally-constrained group sparse learning method aiming for longitudinal analysis with multiple time-points of data. Specifically, for each time-point, we train a sparse linear regression model by using the imaging data and the corresponding responses, and further use the

group regularization

to group the weights corresponding to the same brain region across different time-points together. Moreover, to reflect the smooth changes between adjacent time-points of data, we also include two

smoothness regularization

terms into the objective function, i.e., one

fused smoothness

term which requires the differences between two successive weight vectors from adjacent time-points should be small, and another

output smoothness

term which requires the differences between outputs of two successive models from adjacent time-points should also be small. We develop an efficient algorithm to solve the new objective function with both group-sparsity and smoothness regularizations. We validate our method through estimation of clinical cognitive scores using imaging data at multiple time-points which are available in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database.

Transcranial sonography (TCS) is a new tool for the diagnosis of Parkinson’s disease (PD) according to a distinct hyperechogenic pattern in the substantia nigra (SN) region. However a procedure including rating scale of SN hyperechogenicity was required for a standard clinical setting with increased use. We applied the feature analysis method to a large TCS dataset that is relevant for clinical practice and includes the variability that is present under real conditions. In order to decrease the influence to the image properties from the different settings of ultrasound machine, we propose a local image analysis method using an invariant scale blob detection for the hyperechogenicity estimation. The local features are extracted from the detected blobs and the watershed regions in half of mesencephalon area. The performance of these features is evaluated by a feature-selection method. The cross validation results show that the local features could be used for PD detection.

Longitudinal imaging studies are frequently used to investigate temporal changes in brain morphology. Image intensity may also change over time, for example when studying brain maturation. However, such intensity changes are not accounted for in image similarity measures for standard image registration methods. Hence, (i) local similarity measures, (ii) methods estimating intensity transformations between images, and (iii) metamorphosis approaches have been developed to either achieve robustness with respect to intensity changes or to simultaneously capture spatial and intensity changes. For these methods, longitudinal intensity changes are not explicitly modeled and images are treated as independent static samples. Here, we propose a

model-based image similarity measure

for longitudinal image registration in the presence of spatially non-uniform intensity change.

Cortical thinning is a widely used and powerful biomarker for measuring disease progression in Alzheimer’s disease (AD). However, there has been little work on the effect of atrophy on the cortical folding patterns. In this study, we examined whether the cortical folding could be used as a biomarker of AD. Cortical folding metrics were computed on 678 patients from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort. For each subject, the boundary between grey matter and white matter was extracted using a level set technique. At each point on the boundary two metrics characterising folding, curvedness and shape index, were generated. Joint histograms using these metrics were calculated for five regions of interest (ROIs): frontal, temporal, occipital, and parietal lobes as well as the cingulum. Pixelwise statistical maps were generated from the joint histograms using permutations tests. In each ROI, a significant reduction was observed between controls and AD in areas associated with the sulcal folds, suggesting a sulcal opening associated with neurodegeneration. When comparing to MCI patients, the regions of significance were smaller but overlapping with those regions found comparing controls to AD. It indicates that the differences in cortical folding are progressive and can be detected before formal diagnosis of AD. Our preliminary analysis showed a viable signal in the cortical folding patterns for Alzheimer’s disease that should be explored further.

Quantitative modeling and analysis of structural and functional brain networks based on diffusion tensor imaging (DTI)/functional MRI (fMRI) data has received extensive interest recently. However, the regularity of these structural or functional brain networks across multiple neuroimaging modalities and across individuals is largely unknown. This paper presents a novel approach to infer group-wise consistent brain sub-networks from multimodal DTI/fMRI datasets via multi-view spectral clustering of cortical networks, which were constructed on our recently developed and extensively validated large-scale cortical landmarks. We applied the proposed algorithm on 80 multimodal structural and functional brain networks of 40 healthy subjects, and obtained consistent multimodal brain sub-networks within the group. Our experiments demonstrated that the derived brain sub-networks have improved inter-modality and inter-subject consistency.

The human connectome has recently become a popular research topic in neuroscience, and many new algorithms have been applied to analyze brain networks. In particular, network topology measures from graph theory have been adapted to analyze network efficiency and ‘small-world’ properties. While there has been a surge in the number of papers examining connectivity through graph theory, questions remain about its test-retest reliability (TRT). In particular, the reproducibility of structural connectivity measures has not been assessed. We examined the TRT of global connectivity measures generated from graph theory analyses of 17 young adults who underwent two high-angular resolution diffusion (HARDI) scans approximately 3 months apart. Of the measures assessed, modularity had the highest TRT, and it was stable across a range of sparsities (a thresholding parameter used to define which network edges are retained). These reliability measures underline the need to develop network descriptors that are robust to acquisition parameters.

Diffusion magnetic resonance imaging has been used extensively to probe the white matter in vivo. Typically, the raw diffusion images are used to reconstruct a diffusion tensor image (DTI). The incapacity of DTI to represent crossing fibers leaded to the development of more sophisticated diffusion models. Among them, multi-fiber models represent each fiber bundle independently, allowing the direct extraction of diffusion features for population analysis. However, no method exists to properly register multi-fiber models, seriously limiting their use in group comparisons. This paper presents a registration and atlas construction method for multi-fiber models. The validity of the registration is demonstrated on a dataset of 45 subjects, including both healthy and unhealthy subjects. Morphometry analysis and tract-based statistics are then carried out, proving that multi-fiber models registration is better at detecting white matter local differences than single tensor registration.

A novel framework for robust 3D tracing in Electron Micrographs is presented. The proposed framework is built using ideas from hypergraph diffusion, and achieves two main objectives. Firstly, the approach scales to trace hundreds of targets without noticeable increase in runtime complexity. Secondly, the framework yields flexibility to fuse top down (global cues as hyperedges) and bottom up (local superpixels as nodes) information. Subsequently, a procedure for auto-seeding to initialize the tracing procedure is proposed. The paper concludes with experimental validation on a challenging large scale tracing problem for simultaneously tracing 95 structures, illustrating applicability of the proposed algorithm.

Fluorescently-tagged proteins located on vesicles can fuse with the surface membrane (visualised as a ‘puff’) or undock and return back into the bulk of the cell. Detection and quantitative measurement of these events from time-lapse videos has proven difficult. We propose a novel approach to detect fusion and undocking events by first searching for docked vesicles that ‘disappear’ from the field of view, and then using a diffusion model to classify them as either fusion or undocking events. We can also use the same searching method to identify docking events. We present comparative results against existing algorithms.

For biologists studying the morphology of cells, Electron Microscopy (EM) is the method of choice with its

nm

resolution. However, the time necessary to acquire EM image series is long and often limits both the number and size of samples imaged. This paper presents a strategy for fast imaging and automated selection of regions of interest that significantly accelerates this process. In particular, this strategy involves scanning a tissue sample once, finding regions of interest in which target structures might be located, scanning these regions once again, and iterating the process until only relevant regions of the block face have been scanned repeatedly. For mitochondria and synapses, this approach is shown to produce near equal localization results to current state-of-the art techniques, and does so in almost a tenth of the time.

In low-resource areas, the most common method of tuberculosis (TB) diagnosis is visual identification of rod-shaped TB bacilli in microscopic images of sputum smears. We present an algorithm for automated TB detection using images from digital microscopes such as CellScope [2], a novel, portable device capable of brightfield and fluorescence microscopy. Automated processing on such platforms could save lives by bringing healthcare to rural areas with limited access to laboratory-based diagnostics. Our algorithm applies morphological operations and template matching with a Gaussian kernel to identify candidate TB-objects. We characterize these objects using Hu moments, geometric and photometric features, and histograms of oriented gradients and then perform support vector machine classification. We test our algorithm on a large set of CellScope images (594 images corresponding to 290 patients) from sputum smears collected at clinics in Uganda. Our object-level classification performance is highly accurate, with Average Precision of 89.2%±2.1%. For slide-level classification, our algorithm performs at the level of human readers, demonstrating the potential for making a significant impact on global healthcare.

Extraction of bone contours from radiographs plays an important role in disease diagnosis, pre-operative planning, and treatment analysis. We present a fully automatic method to accurately segment the proximal femur in anteroposterior pelvic radiographs. A number of candidate positions are produced by a global search with a detector. Each is then refined using a statistical shape model together with local detectors for each model point. Both global and local models use Random Forest regression to vote for the optimal positions, leading to robust and accurate results. The performance of the system is evaluated using a set of 519 images. We show that the fully automated system is able to achieve a mean point-to-curve error of less than 1

mm

for 98% of all 519 images. To the best of our knowledge, this is the most accurate automatic method for segmenting the proximal femur in radiographs yet reported.

We provide a fully automatic method of segmenting vertebrae in DXA images. This is of clinical relevance to the diagnosis of osteoporosis by vertebral fracture, and to grading fractures in clinical trials. In order to locate the vertebrae we train detectors for the upper and lower vertebral endplates. Each detector uses random forest regressor voting applied to Haar-like input features. The regressors are applied at a grid of points across the image, and each tree votes for an endplate centre position. Modes in the smoothed vote image are endplate candidates, some of which are the neighbouring vertebrae of the one sought. The ambiguity is resolved by applying geometric constraints to the connections between vertebrae, although there can be some ambiguity about where the sequence starts (e.g. is the lowest vertebra L4 or L5, Fig 2a). The endplate centres are used to initialise a final phase of Active Appearance Model search for a detailed solution. The method is applied to a dataset of 320 DXA images. Accuracy is comparable to manually initialised AAM segmentation in 91% of images, but multiple grade 3 fractures can cause some edge confusion in severely osteoporotic cases.

We present a method for automatic segmentation of high-grade gliomas and their subregions from multi-channel MR images. Besides segmenting the gross tumor, we also differentiate between active cells, necrotic core, and edema. Our discriminative approach is based on decision forests using context-aware spatial features, and integrates a generative model of tissue appearance, by using the probabilities obtained by tissue-specific Gaussian mixture models as additional input for the forest. Our method classifies the individual tissue types simultaneously, which has the potential to simplify the classification task. The approach is computationally efficient and of low model complexity. The validation is performed on a labeled database of 40 multi-channel MR images, including DTI. We assess the effects of using DTI, and varying the amount of training data. Our segmentation results are highly accurate, and compare favorably to the state of the art.

Magnetic resonance (MR) imaging of carotid atherosclerosis biomarkers are increasingly being investigated for the risk assessment of vulnerable plaques. A fast and robust 3D segmentation of the carotid adventitia (AB) and lumen-intima (LIB) boundaries can greatly alleviate the measurement burden of generating quantitative imaging biomarkers in clinical research. In this paper, we propose a novel global optimization-based approach to segment the carotid AB and LIB from 3D T1-weighted black blood MR images, by simultaneously evolving two coupled surfaces with enforcement of anatomical consistency of the AB and LIB. We show that the evolution of two surfaces at each discrete time-frame can be optimized exactly and globally by means of convex relaxation. Our continuous max-flow based algorithm is implemented in GPUs to achieve high computational performance. The experiment results from 16 carotid MR images show that the algorithm obtained high agreement with manual segmentations and achieved high repeatability in segmentation.

Automatic prostate segmentation plays an important role in image guided radiation therapy. However, accurate prostate segmentation in CT images remains as a challenging problem mainly due to three issues: Low image contrast, large prostate motions, and image appearance variations caused by bowel gas. In this paper, a new patient-specific prostate segmentation method is proposed to address these three issues. The main contributions of our method lie in the following aspects: (1) A new patch based representation is designed in the discriminative feature space to effectively distinguish voxels belonging to the prostate and non-prostate regions. (2) The new patch based representation is integrated with a new sparse label propagation framework to segment the prostate, where candidate voxels with low patch similarity can be effectively removed based on sparse representation. (3) An online update mechanism is adopted to capture more patient-specific information from treatment images scanned in previous treatment days. The proposed method has been extensively evaluated on a prostate CT image dataset consisting of 24 patients with 330 images in total. It is also compared with several state-of-the-art prostate segmentation approaches, and experimental results demonstrate that our proposed method can achieve higher segmentation accuracy than other methods under comparison.

We present a two-stage method for effective and efficient detection of one or multiple anatomical landmarks in an arbitrary 3D volume. The first stage of nearest neighbor matching is to roughly estimate the landmark locations. It searches out of 100,000 volumes for the closest to an input volume and then transfers landmark annotations to the input. The second stage of submodular optimization is to refine the landmark locations by running discriminative landmark detectors within the search ranges constrained by the first stage results. Further it coordinates multiple detectors with a search strategy optimized on the fly to reduce the overall computation cost arising in a submodular formulation. We validate the accuracy, speed and robustness of our approach by detecting body regions and landmarks in a dataset of 2,500 CT scans.

The use of classifier-based object detection has found to be a promising approach in medical anatomy detection. In ultrasound images, the detection task is very challenging due to speckle, shadows and low contrast characteristic features. Typical detection algorithms that use purely intensity-based image features with an exhaustive scan of the image (sliding window approach) tend not to perform very well and incur a very high computational cost. The proposed approach in this paper achieves a significant improvement in detection rates while avoiding exhaustive scanning, thereby gaining a large increase in speed. Our approach uses the combination of local features from an intensity image and global features derived from a local phase-based image known as feature symmetry. The proposed approach has been applied to 2384 two-dimensional (2D) fetal ultrasound abdominal images for the detection of the stomach and the umbilical vein. The results presented show that it outperforms prior related work that uses only local or only global features.

We present a new segmentation approach that combines the strengths of label fusion and spectral clustering. The result is an atlas-based segmentation method guided by contour and texture cues in the test image. This offers advantages for datasets with high variability, making the segmentation less prone to registration errors. We achieve the integration by letting the weights of the graph Laplacian depend on image data, as well as atlas-based label priors. The extracted contours are converted to regions, arranged in a hierarchy depending on the strength of the separating boundary. Finally, we construct the segmentation by a region-wise, instead of voxel-wise, voting, increasing the robustness. Our experiments on cardiac MRI show a clear improvement over majority voting and intensity-weighted label fusion.

Currently, multi-organ segmentation (MOS) in abdominal CT can fail to handle clinical patient population with missing organs due to surgical resection. In order to enable the state-of-the-art MOS for these clinically important cases, we propose 1) automatic missing organ detection (MOD) by testing abnormality of post-surgical organ motion and organ-specific intensity homogeneity, and 2) atlas-based MOS of 10 abdominal organs that handles missing organs automatically. The proposed methods are validated with 44 abdominal CT scans including 9 diseased cases with surgical organ resections, resulting in 93.3% accuracy for MOD and improved overall segmentation accuracy by the proposed MOS method when tested on difficult diseased cases.

Multi-atlas segmentation provides a general purpose, fully automated class of techniques for transferring spatial information from an existing dataset (“atlases”) to a previously unseen context (“target”) through image registration. The method used to combine information after registration (“label fusion”) has a substantial impact on the overall accuracy and robustness. In practice, weighted voting techniques have dramatically outperformed algorithms based on statistical fusion (i.e., algorithms that incorporate rater performance into the estimation process — STAPLE). We posit that a critical limitation of statistical techniques (as generally proposed) is that they fail to incorporate intensity seamlessly into the estimation process and models of observation error. Herein, we propose a novel statistical fusion algorithm, Non-Local STAPLE, which merges the STAPLE framework with a non-local means perspective. Non-Local STAPLE (1) seamlessly integrates intensity into the estimation process, (2) provides a theoretically consistent model of multi-atlas observation error, and (3) largely bypasses the need for group-wise unbiased registrations. We demonstrate significant improvements in two empirical multi-atlas experiments.

The recently proposed Sparse Shape Composition (SSC) opens a new avenue for shape prior modeling. Instead of assuming any parametric model of shape statistics, SSC incorporates shape priors on-the-fly by approximating a shape instance (usually derived from appearance cues) by a sparse combination of shapes in a training repository. Theoretically, one can increase the modeling capability of SSC by including as many training shapes in the repository. However, this strategy confronts two limitations in practice.

First

, since SSC involves an iterative sparse optimization at run-time, the more shape instances contained in the repository, the less run-time efficiency SSC has. Therefore, a compact and informative shape dictionary is preferred to a large shape repository.

Second

, in medical imaging applications, training shapes seldom come in one batch. It is very time consuming and sometimes infeasible to re-construct the shape dictionary every time new training shapes appear. In this paper, we propose an online learning method to address these two limitations. Our method starts from constructing an initial shape dictionary using the K-SVD algorithm. When new training shapes come, instead of re-constructing the dictionary from the ground up, we update the existing one using a block-coordinates descent approach. Using the dynamically updated dictionary, sparse shape composition can be gracefully scaled up to model shape priors from a large number of training shapes without sacrificing run-time efficiency. Our method is validated on lung localization in X-Ray and cardiac segmentation in MRI time series. Compared to the original SSC, it shows comparable performance while being significantly more efficient.

Parkinson’s disease (PD) is a neurodegenerative movement disorder caused by decay of dopaminergic cells in the substantia nigra (SN), which are basal ganglia residing within the midbrain area. In the past two decades, transcranial B-mode sonography (TCUS) has emerged as a viable tool in differential diagnosis of PD and recently has been shown to have promising potential as a screening technique for early detection of PD, even before onset of motor symptoms. In TCUS imaging, the degeneration of SN cells becomes visible as bright and hyper-echogenic speckle patches (SNE) in the midbrain. Recent research proposes the usage of 3D ultrasound imaging in order to make the application of the TCUS technique easier and more objective. In this work, for the first time, we propose an automatic 3D SNE detection approach based on random forests, with a novel formulation of SNE probability that relies on visual context and anatomical priors. On a 3D-TCUS dataset of 11 PD patients and 11 healthy controls, we demonstrate that our SNE detection approach yields promising results with a sensitivity and specificity of around

83%

.

Accurate segmentation of prostate in CT images is important in image-guided radiotherapy. However, it is difficult to localize the prostate in CT images due to low image contrast, unpredicted motion and large appearance variations across different treatment days. To address these issues, we propose a sparse representation based classification method to accurately segment the prostate. The main contributions of this paper are: (1) A discriminant dictionary learning technique is proposed to overcome the limitation of the traditional Sparse Representation based Classifier (SRC). (2) Context features are incorporated into SRC to refine the prostate boundary in an iterative scheme. (3) A residue-based linear regression model is trained to increase the classification performance of SRC and extend it from hard classification to soft classification. To segment the prostate, the new treatment image is first rigidly aligned to the planning image space based on the pelvic bones. Then two sets of location-adaptive SRCs along two coordinate directions are applied on the aligned treatment image to produce a probability map, based on which all previously segmented images of the same patient are rigidly aligned onto the new treatment image and majority voting strategy is further adopted to finally segment the prostate in the new treatment image. The proposed method has been evaluated on a CT dataset consisting of 15 patients and 230 CT images. Promising results have been achieved.

This paper presents a novel method for segmenting functional and anatomical structures simultaneously. The proposed method unifies domains of anatomical and functional images (PET-CT), represents them in a product lattice, and performs simultaneous delineation of regions based on a random walk image segmentation. In addition, we propose a simple yet efficient object/background seed localization method, where background and foreground object cues are automatically obtained from PET images and propagated onto the corresponding anatomical images (CT). In our experiments, abnormal anatomies on PET-CT images from human subjects are segmented synergistically by the proposed fully automatic co-segmentation method with high precision (mean DSC of 91.44%) in seconds (avg. 40 seconds).

The paper presents a method for learning multimodal classifiers from datasets in which not all subjects have data from all modalities. Usually, subjects with a severe form of pathology are the ones failing to satisfactorily complete the study, especially when it consists of multiple imaging modalities. A classifier capable of handling subjects with unequal numbers of modalities prevents discarding any subjects, as is traditionally done, thereby broadening the scope of the classifier to more severe pathology. It also allows design of the classifier to include as much of the available information as possible and facilitates testing of subjects with missing modalities over the constructed classifier. The presented method employs an ensemble based approach where several subsets of complete data are formed and trained using individual classifiers. The output from these classifiers is fused using a weighted aggregation step giving an optimal probabilistic score for each subject. The method is applied to a spatio-temporal dataset for autism spectrum disorders (ASD)(96 patients with ASD and 42 typically developing controls) that consists of functional features from magnetoencephalography (MEG) and structural connectivity features from diffusion tensor imaging (DTI). A clear distinction between ASD and controls is obtained with an average 5-fold accuracy of 83.3% and testing accuracy of 88.4%. The fusion classifier performance is superior to the classification achieved using single modalities as well as multimodal classifier using only complete data (78.3%). The presented multimodal classifier framework is applicable to all modality combinations.

Diffusion imaging, through the study of water diffusion, allows for the characterization of brain white matter, both at the population and individual level. In recent years, it has been employed to detect brain abnormalities in patients suffering from a disease, e.g. from multiple sclerosis (MS). State-of-the-art methods usually utilize a database of matched (age, sex, ...) controls, registered onto a template, to test for differences in the patient white matter. Such approaches however suffer from two main drawbacks. First, registration algorithms are prone to local errors, thereby degrading the comparison results. Second, the database needs to be large enough to obtain reliable results. However, in medical imaging, such large databases are hardly available. In this paper, we propose a new method that addresses these two issues. It relies on the search for samples in a local neighborhood of each pixel to increase the size of the database. Then, we propose a new test based on these samples to perform a voxelwise comparison of a patient image with respect to a population of controls. We demonstrate on simulated and real MS patient data how such a framework allows for an improved detection power and a better robustness and reproducibility, even with a small database.

Fiber clustering is an essential step towards brain connectivity modeling and tract-based analysis of white matter integrity via diffusion tensor imaging (DTI) in many clinical neuroscience applications. A variety of methods have been developed to cluster fibers based on various types of features such as geometry, anatomy, connection, or function. However, identification of group-wise consistent fiber bundles that are harmonious across multi-modalities is rarely explored yet. This paper proposes a novel hybrid two-stage approach that incorporates connectional and functional features, and identifies group-wise consistent fiber bundles across subjects. In the first stage, based on our recently developed 358 dense and consistent cortical landmarks, we identified consistent backbone bundles with representative fibers. In the second stage, other remaining fibers are then classified into the existing backbone bundles using their correlations of resting state fMRI signals at the two ends of fibers. Our experimental results show that the proposed methods can achieve group-wise consistent fiber bundles with similar shapes and anatomic profiles, as well as strong functional coherences.

Having to determine an adequate number of fiber directions is a fundamental limitation of multi-compartment models in diffusion MRI. This paper proposes a novel strategy to approach this problem, based on simulating data that closely follows the characteristics of the measured data. This provides the ground truth required to determine the number of directions that optimizes a formal measure of accuracy, while allowing us to transfer the result to real data by support vector regression. The method is shown to result in plausible and reproducible decisions on three repeated scans of the same subject. When combined with the ball-and-stick model, it produces directional estimates comparable to constrained spherical deconvolution, but with significantly smaller variance between re-scans, and at a reduced computational cost.

In clinical practice, physicians often exploit previously observed patterns in coronary angiograms from similar patients to quickly assess the state of the disease in a current patient. These assessments involve visually observed features such as the distance of a junction from the root and the tortuosity of the arteries. In this paper, we show how these visual features can be automatically extracted from coronary artery images and used for finding similar coronary angiograms from a database. Testing on a large collection has shown the method finds clinically similar coronary angiograms from patients with similar clinical history.

Assessment of trauma patients with multiple injuries can be one of the most clinically challenging situations dealt with by the radiologist. We propose a fully automated method to detect acute vertebral body fractures on trauma CT studies. The spine is first segmented and partitioned into vertebrae. Then the cortical shell of the vertebral body is extracted using deformable dual-surface models. The extracted cortical shell is unwrapped onto a 2D map effectively converting a complex 3D fracture detection problem into a pattern recognition problem of fracture lines on a 2D plane. Twenty-eight features are computed for each fracture line and sent to a committee of support vector machines for classification. The system was tested on 18 trauma CT datasets and achieved 95.3% sensitivity and 1.7 false positives per case by leave-one-out cross validation.

Texture–based computerized analysis of high–resolution computed tomography images from patients with interstitial lung diseases is introduced to assist radiologists in image interpretation. The cornerstone of our approach is to learn lung texture signatures using a linear combination of

N

–th order Riesz templates at multiple scales. The weights of the linear combination are derived from one–versus–all support vector machines. Steerability and multiscale properties of Riesz wavelets allow for scale and rotation covariance of the texture descriptors with infinitesimal precision. Orientations are normalized among texture instances by locally aligning the Riesz templates, which is carried out analytically. The proposed approach is compared with state–of–the–art texture attributes and shows significant improvement in classification performance with an average area under receiver operating characteristic curves of 0.94 for five lung tissue classes. The derived lung texture signatures illustrate optimal class–wise discriminative properties.

In this paper we present a hybrid 0D-3D modeling method to investigate the hepatic flow in a virtual right lobe hepatectomy (RLH), the surgical procedure for adult-to-adult living donor liver transplanation (LDLT). The 3D method is employed to simulate complex 3D flow in the portal vein, and the 0D model is used to study the systemic hepatic circulation. In particular, we quantify the flow velocity and wall shear stress (WSS) in the left portal vein which increase dramatically post-RLH, and also simulate the essential hepatic distribution features in a healthy adult pre- and post-procedure. We further predict the arterial flow in the remnant left liver, which would decrease due to a hepatic arterial buffer response (HABR) effect. Finally we discuss the physiological significance of these phenomena, and the potential of this hybrid modeling approach.

We present a new method to analyze, classify and characterize 3D landmark-based shapes. It is based on a framework provided by oriented matroid theory, that is on a combinatorial encoding of convexity properties. We apply this method to a set of skull shapes presenting various types of coronal craniosynostosis.

Arterial Spin Labeling (ASL) enables measuring cerebral blood flow in MRI without injection of a contrast agent. Perfusion measured by ASL carries relevant information for patients suffering from pathologies associated with singular perfusion patterns. However, to date, individual identification of abnormal perfusion patterns in ASL usually relies on visual inspection or manual delineation of regions of interest.

In this paper, we introduce a new framework to automatically outline patterns of abnormal perfusion in individual patients by means of an ASL template. We compare two models of normal perfusion and assess the quality of detections comparing an

a contrario

approach to the Generalized Linear Model (GLM).

In this study, we propose a computational diagnosis system for detecting the colorectal cancer from histopathological slices.The computational analysis was usually performed on patch level where only a small part of the slice is covered. However, slice-based classification is more realistic for histopathological diagnosis. The developed method combines both textural and structural features from patch images and proposes a two level classification scheme. In the first level, the patches in slices are classified into possible classes (adenomatous, inflamed, cancer and normal) and the distribution of the patches into these classes is considered as the information representing the slices. Then the slices are classified using a logistic linear classifier. In patch level, we obtain the correct classification accuracies of 94.36% and 96.34% for the cancer and normal classes, respectively. However, in slice level, the accuracies of the 79.17% and 92.68% are achieved for cancer and normal classes, respectively.

Dynamic PET imaging provides important spatial-temporal information for metabolism analysis of organs and tissues, and generates a great reference for clinical diagnosis and pharmacokinetic analysis. Due to poor statistical properties of the measurement data in low count dynamic PET acquisition and disturbances from surrounding tissues, identifying small lesions inside the human body is still a challenging issue. The uncertainties in estimating the arterial input function will also limit the accuracy and reliability of the metabolism analysis of lesions. Furthermore, the sizes of the patients and the motions during PET acquisition will yield mismatch against general purpose reconstruction system matrix, this will also affect the quantitative accuracy of metabolism analyses of lesions. In this paper, we present a dynamic PET metabolism analysis framework by defining a patient adaptive system matrix to improve the lesion metabolism analysis. Both patient size information and potential small lesions are incorporated by simulations of phantoms of different sizes and individual point source responses. The new framework improves the quantitative accuracy of lesion metabolism analysis, and makes the lesion identification more precisely. The requirement of accurate input functions is also reduced. Experiments are conducted on Monte Carlo simulated data set for quantitative analysis and validation, and on real patient scans for assessment of clinical potential.

Time-resolved imaging of the thorax or abdominal area is affected by respiratory motion. Nowadays, one-dimensional respiratory surrogates are used to estimate the current state of the lung during its cycle, but with rather poor results. This paper presents a framework to predict the 3D lung motion based on a patient-specific finite element model of respiratory mechanics estimated from two CT images at end of inspiration (EI) and end of expiration (EE). We first segment the lung, thorax and sub-diaphragm organs automatically using a machine-learning algorithm. Then, a biomechanical model of the lung, thorax and sub-diaphragm is employed to compute the 3D respiratory motion. Our model is driven by thoracic pressures, estimated automatically from the EE and EI images using a trust-region approach. Finally, lung motion is predicted by modulating the thoracic pressures. The effectiveness of our approach is evaluated by predicting lung deformation during exhale on five DIR-Lab datasets. Several personalization strategies are tested, showing that an average error of 3.88 ±1.54

mm

in predicted landmark positions can be achieved. Since our approach is generative, it may constitute a 3D surrogate information for more accurate medical image reconstruction and patient respiratory analysis.

Distortion correction is applied to endoscopic duodenal imagery to improve automated classification of celiac disease affected mucosa patches. In a set of six edge- and shape-related feature extraction techniques, only a single one is able to consistently benefit from distortion correction, while for others, even a decrease of classification accuracy is observed. Different types of distortion correction do not lead to significantly different behaviour in the observed application scenario.

In this paper, we propose a new pharmacokinetic model for parameter estimation of dynamic contrast-enhanced (DCE) MRI by using Gaussian process inference. Our model is based on the Tofts dual-compartment model for the description of tracer kinetics and the observed time series from DCE-MRI is treated as a Gaussian stochastic process. The parameter estimation is done through a maximum likelihood approach and we propose a variant of the coordinate descent method to solve this likelihood maximization problem. The new model was shown to outperform a baseline method on simulated data. Parametric maps generated on prostate DCE data with the new model also provided better enhancement of tumors, lower intensity on false positives, and better boundary delineation when compared with the baseline method. New statistical parameter maps from the process model were also found to be informative, particularly when paired with the PK parameter maps.

This paper presents a new method for automatic localization and identification of vertebrae in arbitrary field-of-view CT scans. No assumptions are made about which section of the spine is visible or to which extent. Thus, our approach is more general than previous work while being computationally efficient. Our algorithm is based on regression forests and probabilistic graphical models. The discriminative, regression part aims at roughly detecting the visible part of the spine. Accurate localization and identification of individual vertebrae is achieved through a generative model capturing spinal shape and appearance. The system is evaluated quantitatively on 200 CT scans, the largest dataset reported for this purpose. We obtain an overall median localization error of less than 6mm, with an identification rate of 81%.

With improvements in acquisition speed and quality, the amount of medical image data to be screened by clinicians is starting to become challenging in the daily clinical practice. To quickly visualize and find abnormalities in medical images, we propose a new method combining segmentation algorithms with statistical shape models. A statistical shape model built from a healthy population will have a close fit in healthy regions. The model will however not fit to morphological abnormalities often present in the areas of pathologies. Using the residual fitting error of the statistical shape model, pathologies can be visualized very quickly. This idea is applied to finding drusen in the retinal pigment epithelium (RPE) of optical coherence tomography (OCT) volumes. A segmentation technique able to accurately segment drusen in patients with age-related macular degeneration (AMD) is applied. The segmentation is then analyzed with a statistical shape model to visualize potentially pathological areas. An extensive evaluation is performed to validate the segmentation algorithm, as well as the quality and sensitivity of the hinting system. Most of the drusen with a height of 85.5

μ

m were detected, and all drusen at least 93.6

μ

m high were detected.

Analyzing geometry of sulcal curves on the human cortical surface requires a shape representation invariant to Euclidean motion. We present a novel shape representation that characterizes the shape of a curve in terms of a coordinate system based on the eigensystem of the anisotropic Helmholtz equation. This representation has many desirable properties: stability, uniqueness and invariance to scaling and isometric transformation. Under this representation, we can find a point-wise shape distance between curves as well as a bijective smooth point-to-point correspondence. When the curves are sampled irregularly, we also present a fast and accurate computational method for solving the eigensystem using a finite element formulation. This shape representation is used to find symmetries between corresponding sulcal shapes between cortical hemispheres. For this purpose, we automatically generate 26 sulcal curves for 24 subject brains and then compute their invariant shape representation. Left-right sulcal shape symmetry as measured by the shape representation’s metric demonstrates the utility of the presented invariant representation for shape analysis of the cortical folding pattern.

The restoration of microscopy images makes the segmentation and detection of cells easier and more reliable, which facilitates automated cell tracking and cell behavior analysis. In this paper, the authors analyze the image formation process of phase contrast images and propose an image restoration method based on the dictionary representation of diffraction patterns. By formulating and solving a

min

-ℓ

1

optimization problem, each pixel is restored into a feature vector corresponding to the dictionary representation. Cells in the images are then segmented by the feature vector clustering. In addition to segmentation, since the feature vectors capture the information on the phase retardation caused by cells, they can be used for cell stage classification between intermitotic and mitotic/apoptotic stages. Experiments on three image sequences demonstrate that the dictionary-based restoration method can restore phase contrast images containing cells with different optical natures and provide promising results on cell stage classification.

Histopathology image segmentation plays a very important role in cancer diagnosis and therapeutic treatment. Existing supervised approaches for image segmentation require a large amount of high quality manual delineations (on pixels), which is often hard to obtain. In this paper, we propose a new algorithm along the line of weakly supervised learning; we introduce context constraints as a prior for multiple instance learning (ccMIL), which significantly reduces the ambiguity in weak supervision (a 20% gain); our method utilizes image-level labels to learn an integrated model to perform histopathology cancer image segmentation, clustering, and classification. Experimental results on colon histopathology images demonstrate the great advantages of ccMIL.

We cast segmentation of 3D tubular structures in a bundle as partitioning of

structural-flow trajectories

. Traditional 3D segmentation algorithms aggregate local pixel correlations incrementally along a 3D stack. In contrast, structural-flow trajectories establish

long range

pixel correspondences and their affinities propagate grouping cues across the entire volume simultaneously, from informative to non-informative places. Segmentation by trajectory clustring recovers from persistent ambiguities caused by faint boundaries or low contrast, common in medical images. Trajectories are computed by linking successive registration fields, each one registering pairs of consecutive slices of the 3D stack. We show our method effectively unravels densely packed tubular structures, without any supervision or 3D shape priors, outperforming previous 2D and 3D segmentation algorithms.

We present an original method for the online blind calibration of non-uniform photodetectors. The disparity of the detectors may arise from both irregular spatial arrangement and distinct slowly time-varying photometric transfer functions. As natural images are mostly continuous, the signal

collected

by neighboring detectors is strongly correlated over time. The core idea of our method is to translate the calibration problem into relative pairwise calibrations between neighboring detectors followed by the regularized inversion of a system akin to gradient-based surface recovery. From our blind calibration procedure, we design an

online

blind calibration pipeline compatible with clinical practice.

Online

blind

calibration is proved to be statistically better than standard

offline

calibration for reconstructing endomicroscopy sequences.