From Local to Global: A Graph Framework for Retinal Artery/Vein Classification

Fundus photography has been widely used for inspecting eye disorders by ophthalmologists or computer algorithms. Biomarkers related to retinal vessels plays an essential role to detect early diabetes. To quantify vascular biomarkers or the corresponding changes, an accurate artery and vein classification is necessary. In this work, we propose a new framework to boost local vessel classification with a global vascular network model using graph convolution. We compare our proposed method with two traditional state-of-the-art methods on a testing dataset of 750 images from the Maastricht Study. After incorporating global information, our model achieves the best accuracy of 86.45% compared to 85.5% from convolutional neural networks (CNN) and 82.9% from handcrafted pixel feature classification (HPFC). Our model also obtains the best area under receiver operating characteristic curve (AUC) of 0.95, compared to 0.93 from CNN and 0.90 from HPFC. The new classification framework has the advantage of easy deployment on top of local classification features. It corrects the local classification error by minimizing global classification error and it brings free additional classification performance.

Type 2 diabetes and HbA1c are independently associated with wider retinal arterioles: the Maastricht study

AIMS/HYPOTHESIS

Retinal microvascular diameters are biomarkers of cardio-metabolic risk. However, the association of (pre)diabetes with retinal microvascular diameters remains unclear. We aimed to investigate the association of prediabetes (impaired fasting glucose or impaired glucose tolerance) and type 2 diabetes with retinal microvascular diameters in a predominantly white population.

METHODS

In a population-based cohort study with oversampling of type 2 diabetes (N = 2876; n = 1630 normal glucose metabolism [NGM], n = 433 prediabetes and n = 813 type 2 diabetes, 51.2% men, aged 59.8 ± 8.2 years; 98.6% white), we determined retinal microvascular diameters (measurement unit as measured by retinal health information and notification system [RHINO] software) and glucose metabolism status (using OGTT). Associations were assessed with multivariable regression analyses adjusted for age, sex, waist circumference, smoking, systolic blood pressure, lipid profile and the use of lipid-modifying and/or antihypertensive medication.

RESULTS

Multivariable regression analyses showed a significant association for type 2 diabetes but not for prediabetes with arteriolar width (vs NGM; prediabetes: β = 0.62 [95%CI -1.58, 2.83]; type 2 diabetes: 2.89 [0.69, 5.08]; measurement unit); however, there was a linear trend for the arteriolar width across glucose metabolism status (p for trend = 0.013). The association with wider venules was not statistically significant (prediabetes: 2.40 [-1.03, 5.84]; type 2 diabetes: 2.87 [-0.55, 6.29], p for trend = 0.083; measurement unit). Higher HbA_1c levels were associated with wider retinal arterioles (standardised β = 0.043 [95% CI 0.00002, 0.085]; p = 0.050) but the association with wider venules did not reach statistical significance (0.037 [-0.006, 0.080]; p = 0.092) after adjustment for potential confounders.

CONCLUSIONS/INTERPRETATION

Type 2 diabetes, higher levels of HbA_1c and, possibly, prediabetes, are independently associated with wider retinal arterioles in a predominantly white population. These findings indicate that microvascular dysfunction is an early phenomenon in impaired glucose metabolism.

Reconnection of Interrupted Curvilinear Structures via Cortically Inspired Completion for Ophthalmologic Images

OBJECTIVE

In this paper, we propose a robust, efficient, and automatic reconnection algorithm for bridging interrupted curvilinear skeletons in ophthalmologic images.

METHODS

This method employs the contour completion process, i.e., mathematical modeling of the direction process in the roto-translation group to achieve line propagation/completion. The completion process can be used to reconstruct interrupted curves by considering their local consistency. An explicit scheme with finite-difference approximation is used to construct the three-dimensional (3-D) completion kernel, where we choose the Gamma distribution for time integration. To process structures in , the orientation score framework is exploited to lift the 2-D curvilinear segments into the 3-D space. The propagation and reconnection of interrupted segments are achieved by convolving the completion kernel with orientation scores via iterative group convolutions. To overcome the problem of incorrect skeletonization of 2-D structures at junctions, a 3-D segment-wise thinning technique is proposed to process each segment separately in orientation scores.

RESULTS

Validations on 4 datasets with different image modalities show that our method achieves an average success rate of in reconnecting gaps of sizes from to , including challenging junction structures.

CONCLUSION

The reconnection approach can be a useful and reliable technique for bridging complex curvilinear interruptions.

SIGNIFICANCE

The presented method is a critical work to obtain more complete curvilinear structures in ophthalmologic images. It provides better topological and geometric connectivities for further analysis.

Retinal Microaneurysms Detection Using Local Convergence Index Features

Retinal microaneurysms (MAs) are the earliest clinical sign of diabetic retinopathy disease. Detection of MAs is crucial for the early diagnosis of diabetic retinopathy and prevention of blindness. In this paper, a novel and reliable method for automatic detection of MAs in retinal images is proposed. In the first stage of the proposed method, several preliminary microaneurysm candidates are extracted using a gradient weighting technique and an iterative thresholding approach. In the next stage, in addition to intensity and shape descriptors, a new set of features based on local convergence index filters is extracted for each candidate. Finally, the collective set of features is fed to a hybrid sampling/boosting classifier to discriminate the MAs from non-MAs candidates. The method is evaluated on images with different resolutions and modalities (color and scanning laser ophthalmoscope) using six publicly available data sets including the retinopathy online challenges (ROC) data set. The proposed method achieves an average sensitivity score of 0.471 on the ROC data set outperforming state-of-the-art approaches in an extensive comparison. The experimental results on the other five data sets demonstrate the effectiveness and robustness of the proposed MAs detection method regardless of different image resolutions and modalities.

Retinal artery/vein classification using genetic-search feature selection

BACKGROUND AND OBJECTIVES

The automatic classification of retinal blood vessels into artery and vein (A/V) is still a challenging task in retinal image analysis. Recent works on A/V classification mainly focus on the graph analysis of the retinal vasculature, which exploits the connectivity of vessels to improve the classification performance. While they have overlooked the importance of pixel-wise classification to the final classification results. This paper shows that a complicated feature set is efficient for vessel centerline pixels classification.

METHODS

We extract enormous amount of features for vessel centerline pixels, and apply a genetic-search based feature selection technique to obtain the optimal feature subset for A/V classification.

RESULTS

The proposed method achieves an accuracy of 90.2%, the sensitivity of 89.6%, the specificity of 91.3% on the INSPIRE dataset. It shows that our method, using only the information of centerline pixels, gives a comparable performance as the techniques which use complicated graph analysis. In addition, the results on the images acquired by different fundus cameras show that our framework is capable for discriminating vessels independent of the imaging device characteristics, image resolution and image quality.

CONCLUSION

The complicated feature set is essential for A/V classification, especially on the individual vessels where graph-based methods receive limitations. And it could provide a higher entry to the graph-analysis to achieve a better A/V labeling.

Curvature Integration in a 5D Kernel for Extracting Vessel Connections in Retinal Images

Tree-like structures, such as retinal images, are widely studied in computer-aided diagnosis systems for large-scale screening programs. Despite several segmentation and tracking methods proposed in the literature, there still exist several limitations specifically when two or more curvilinear structures cross or bifurcate, or in the presence of interrupted lines or highly curved blood vessels. In this paper, we propose a novel approach based on multi-orientation scores augmented with a contextual affinity matrix, which both are inspired by the geometry of the primary visual cortex (V1) and their contextual connections. The connectivity is described with a 5D kernel obtained as the fundamental solution of the Fokker-Planck equation modeling the cortical connectivity in the lifted space of positions, orientations, curvatures, and intensity. It is further used in a self-tuning spectral clustering step to identify the main perceptual units in the stimuli. The proposed method has been validated on several easy as well as challenging structures in a set of artificial images and actual retinal patches. Supported by quantitative and qualitative results, the method is capable of overcoming the limitations of current state-of-the-art techniques.

Template Matching via Densities on the Roto-Translation Group

We propose a template matching method for the detection of 2D image objects that are characterized by orientation patterns. Our method is based on data representations via orientation scores, which are functions on the space of positions and orientations, and which are obtained via a wavelet-type transform. This new representation allows us to detect orientation patterns in an intuitive and direct way, namely via cross-correlations. Additionally, we propose a generalized linear regression framework for the construction of suitable templates using smoothing splines. Here, it is important to recognize a curved geometry on the position-orientation domain, which we identify with the Lie group SE(2): the roto-translation group. Templates are then optimized in a B-spline basis, and smoothness is defined with respect to the curved geometry. We achieve state-of-the-art results on three different applications: detection of the optic nerve head in the retina (99.83 percent success rate on 1,737 images), of the fovea in the retina (99.32 percent success rate on 1,616 images), and of the pupil in regular camera images (95.86 percent on 1,521 images). The high performance is due to inclusion of both intensity and orientation features with effective geometric priors in the template matching. Moreover, our method is fast due to a cross-correlation based matching approach.

Robust Retinal Vessel Segmentation via Locally Adaptive Derivative Frames in Orientation Scores

This paper presents a robust and fully automatic filter-based approach for retinal vessel segmentation. We propose new filters based on 3D rotating frames in so-called orientation scores, which are functions on the Lie-group domain of positions and orientations [Formula: see text]. By means of a wavelet-type transform, a 2D image is lifted to a 3D orientation score, where elongated structures are disentangled into their corresponding orientation planes. In the lifted domain [Formula: see text], vessels are enhanced by means of multi-scale second-order Gaussian derivatives perpendicular to the line structures. More precisely, we use a left-invariant rotating derivative (LID) frame, and a locally adaptive derivative (LAD) frame. The LAD is adaptive to the local line structures and is found by eigensystem analysis of the left-invariant Hessian matrix (computed with the LID). After multi-scale filtering via the LID or LAD in the orientation score domain, the results are projected back to the 2D image plane giving us the enhanced vessels. Then a binary segmentation is obtained through thresholding. The proposed methods are validated on six retinal image datasets with different image types, on which competitive segmentation performances are achieved. In particular, the proposed algorithm of applying the LAD filter on orientation scores (LAD-OS) outperforms most of the state-of-the-art methods. The LAD-OS is capable of dealing with typically difficult cases like crossings, central arterial reflex, closely parallel and tiny vessels. The high computational speed of the proposed methods allows processing of large datasets in a screening setting.

Ultra-High Field MRI Post Mortem Structural Connectivity of the Human Subthalamic Nucleus, Substantia Nigra, and Globus Pallidus

Introduction: The subthalamic nucleus, substantia nigra, and globus pallidus, three nuclei of the human basal ganglia, play an important role in motor, associative, and limbic processing. The network of the basal ganglia is generally characterized by a direct, indirect, and hyperdirect pathway. This study aims to investigate the mesoscopic nature of these connections between the subthalamic nucleus, substantia nigra, and globus pallidus and their surrounding structures.

Methods: A human post mortem brain specimen including the substantia nigra, subthalamic nucleus, and globus pallidus was scanned on a 7 T MRI scanner. High resolution diffusion weighted images were used to reconstruct the fibers intersecting the substantia nigra, subthalamic nucleus, and globus pallidus. The course and density of these tracks was analyzed.

Results: Most of the commonly established projections of the subthalamic nucleus, substantia nigra, and globus pallidus were successfully reconstructed. However, some of the reconstructed fiber tracks such as the connections of the substantia nigra pars compacta to the other included nuclei and the connections with the anterior commissure have not been shown previously. In addition, the quantitative tractography approach showed a typical degree of connectivity previously not documented. An example is the relatively larger projections of the subthalamic nucleus to the substantia nigra pars reticulata when compared to the projections to the globus pallidus internus.

Discussion: This study shows that ultra-high field post mortem tractography allows for detailed 3D reconstruction of the projections of deep brain structures in humans. Although the results should be interpreted carefully, the newly identified connections contribute to our understanding of the basal ganglia.

Structural Brain Network: What is the Effect of LiFE Optimization of Whole Brain Tractography?

Structural brain networks constructed based on diffusion-weighted MRI (dMRI) have provided a systems perspective to explore the organization of the human brain. Some redundant and nonexistent fibers, however, are inevitably generated in whole brain tractography. We propose to add one critical step while constructing the networks to remove these fibers using the linear fascicle evaluation (LiFE) method, and study the differences between the networks with and without LiFE optimization. For a cohort of nine healthy adults and for 9 out of the 35 subjects from Human Connectome Project, the T 1-weighted images and dMRI data are analyzed. Each brain is parcellated into 90 regions-of-interest, whilst a probabilistic tractography algorithm is applied to generate the original connectome. The elimination of redundant and nonexistent fibers from the original connectome by LiFE creates the optimized connectome, and the random selection of the same number of fibers as the optimized connectome creates the non-optimized connectome. The combination of parcellations and these connectomes leads to the optimized and non-optimized networks, respectively. The optimized networks are constructed with six weighting schemes, and the correlations of different weighting methods are analyzed. The fiber length distributions of the non-optimized and optimized connectomes are compared. The optimized and non-optimized networks are compared with regard to edges, nodes and networks, within a sparsity range of 0.75-0.95. It has been found that relatively more short fibers exist in the optimized connectome. About 24.0% edges of the optimized network are significantly different from those in the non-optimized network at a sparsity of 0.75. About 13.2% of edges are classified as false positives or the possible missing edges. The strength and betweenness centrality of some nodes are significantly different for the non-optimized and optimized networks, but not the node efficiency. The normalized clustering coefficient, the normalized characteristic path length and the small-worldness are higher in the optimized network weighted by the fiber number than in the non-optimized network. These observed differences suggest that LiFE optimization can be a crucial step for the construction of more reasonable and more accurate structural brain networks.

The influence of construction methodology on structural brain network measures: A review

Structural brain networks based on diffusion MRI and tractography show robust attributes such as small-worldness, hierarchical modularity, and rich-club organization. However, there are large discrepancies in the reports about specific network measures. It is hypothesized that these discrepancies result from the influence of construction methodology. We surveyed the methodological options and their influences on network measures. It is found that most network measures are sensitive to the scale of brain parcellation, MRI gradient schemes and orientation model, and the tractography algorithm, which is in accordance with the theoretical analysis of the small-world network model. Different network weighting schemes represent different attributes of brain networks, which makes these schemes incomparable between studies. Methodology choice depends on the specific study objectives and a clear understanding of the pros and cons of a particular methodology. Because there is no way to eliminate these influences, it seems more practical to quantify them, optimize the methodologies, and construct structural brain networks with multiple spatial resolutions, multiple edge densities, and multiple weighting schemes.

Magnetic resonance imaging techniques for visualization of the subthalamic nucleus

The authors reviewed 70 publications on MR imaging-based targeting techniques for identifying the subthalamic nucleus (STN) for deep brain stimulation in patients with Parkinson disease. Of these 70 publications, 33 presented quantitatively validated results. There is still no consensus on which targeting technique to use for surgery planning; methods vary greatly between centers. Some groups apply indirect methods involving anatomical landmarks, or atlases incorporating anatomical or functional data. Others perform direct visualization on MR imaging, using T2-weighted spin echo or inversion recovery protocols. The combined studies do not offer a straightforward conclusion on the best targeting protocol. Indirect methods are not patient specific, leading to varying results between cases. On the other hand, direct targeting on MR imaging suffers from lack of contrast within the subthalamic region, resulting in a poor delineation of the STN. These deficiencies result in a need for intraoperative adaptation of the original target based on test stimulation with or without microelectrode recording. It is expected that future advances in MR imaging technology will lead to improvements in direct targeting. The use of new MR imaging modalities such as diffusion MR imaging might even lead to the specific identification of the different functional parts of the STN, such as the dorsolateral sensorimotor part, the target for deep brain stimulation.