Qiangfeng Peter Lau

Retinal Vascular Tortuosity, Blood Pressure, and Cardiovascular Risk Factors

OBJECTIVE To examine the relationship of retinal vascular tortuosity to age, blood pressure, and other cardiovascular risk factors. DESIGN Population-based, cross-sectional study. PARTICIPANTS A total of 3280 participants aged 40 to 80 years from the Singapore Malay Eye Study (78.7% response rate). METHODS Retinal arteriolar and venular (vascular) tortuosity were quantitatively measured from fundus images using a computer-assisted program. Retinal vascular tortuosity was defined as the integral of the curvature square along the path of the vessel, normalized by the total path length. Data on blood pressure and major cardiovascular disease (CVD) risk factors were collected from all participants. MEAN OUTCOME MEASURES Retinal arteriolar and venular tortuosity. RESULTS A total of 2915 participants contributed data to this study. The mean (standard deviation) and median were 2.99 (1.40) and 2.73 for retinal arteriolar tortuosity (×10(4)), and 4.64 (2.39) and 4.19 for retinal venular tortuosity (×10(4)), respectively. Retinal venules were significantly more tortuous than retinal arterioles (P<0.001). In multivariable-adjusted linear regression models, less arteriolar tortuosity was independently associated with older age, higher blood pressure, higher body mass index (BMI), and narrower retinal arteriolar caliber (all P<0.05); greater venular tortuosity was independently associated with younger age, higher blood pressure, lower high-density lipoprotein (HDL) cholesterol level, and wider retinal venular caliber (all P<0.05). CONCLUSIONS Retinal arteriolar tortuosity was associated with older age and higher levels of blood pressure and BMI, whereas venular tortuosity was also associated with lower HDL level. The quantitative assessment of retinal vascular tortuosity from retinal images may provide further information regarding effects of cardiovascular risk factors on the retinal vasculature.

A New Method to Measure Peripheral Retinal Vascular Caliber over an Extended Area

Please cite this paper as: Cheung, Hsu, Lee, Wang, Mitchell, Lau, Hamzah, Ho and Wong (2010). A New Method to Measure Peripheral Retinal Vascular Caliber over an Extended Area. Microcirculation17(7), 495–503.

Simultaneously Identifying All True Vessels From Segmented Retinal Images

Measurements of retinal blood vessel morphology have been shown to be related to the risk of cardiovascular diseases. The wrong identification of vessels may result in a large variation of these measurements, leading to a wrong clinical diagnosis. In this paper, we address the problem of automatically identifying true vessels as a postprocessing step to vascular structure segmentation. We model the segmented vascular structure as a vessel segment graph and formulate the problem of identifying vessels as one of finding the optimal forest in the graph given a set of constraints. We design a method to solve this optimization problem and evaluate it on a large real-world dataset of 2446 retinal images. Experiment results are analyzed with respect to actual measurements of vessel morphology. The results show that the proposed approach is able to achieve 98.9% pixel precision and 98.7% recall of the true vessels for clean segmented retinal images, and remains robust even when the segmented image is noisy.

Retinal Vascular Fractal Dimension and Its Relationship With Cardiovascular and Ocular Risk Factors

PURPOSE To examine the influence of a range of cardiovascular risk factors and ocular conditions on retinal vascular fractal dimension in the Singapore Malay Eye Study. DESIGN Population-based cross-sectional study. METHODS Fractal analysis of the retinal vessels is a method to quantify the global geometric complexity of the retinal vasculature. Retinal vascular fractal dimension (D(f)) and caliber were measured from retinal photographs using a computer-assisted program. D(f) and arteriolar caliber were combined to form a retinal vascular optimality score (ranging from 0 to 3). Data on cardiovascular and ocular factors were collected from all participants based on a standardized protocol. RESULTS Two thousand nine hundred thirteen (88.8% of 3280 participants) persons had retinal photographs of sufficient quality for the measurement. The mean D(f) was 1.405 (standard deviation, 0.046; interquartile range, 1.243 to 1.542). In the multiple linear regression analysis, after controlling for gender, serum glucose, intraocular pressure, anterior chamber depth, and retinal vascular caliber, smaller D(f) was associated independently with older age (standardized regression coefficient [sβ] = -0.311; P < .001), higher mean arterial blood pressure (sβ = -0.085; P < .001), a more myopic spherical equivalent (sβ = 0.152; P < .001), and presence of cataract (sβ = -0.107; P < .001). Retinal vascular optimality score was associated significantly with higher mean arterial blood pressure (P > .001 for trend). CONCLUSIONS Age, blood pressure, refractive error, and lens opacity had significant influence on retinal vascular fractal measurements. A new score of retinal vascular optimality combining fractals and caliber showed strong association with blood pressure. Quantitative analysis of retinal vasculature therefore may provide additional information on microvascular architecture and optimality.

Retinal Vascular Fractal Dimension Measurement and Its Influence from Imaging Variation: Results of Two Segmentation Methods

Aim: To assess the influences of imaging variation (different photographic angle) on the measurement of retinal vascular fractal dimension (Df), using two segmentation methods. Materials and methods: Nonlinear orthogonal projection segmentation (International Retinal Imaging Software-Fractal, termed IRIS-Fractal) and curvature-based segmentation (Singapore Institute Vessel Assessment-Fractal, termed SIVA-Fractal) methods were used to measure Df and were assessed for their reproducibility in detecting retinal vessels of 30 stereoscopic pairs of optic disc color images. Each pair was taken from the same eye with slightly different angles of incidence. Each photograph of the pairs had subtle variations in brightness between areas temporal and nasal to the optic disc. Results: Intragrader reproducibility of Df measurement was similar (intraclass correlation 0.81 and 0.96, respectively) for IRIS-Fractal and SIVA-Fractal. Within-image pair Pearson’s correlation coefficients (r) of Df measurements were moderate for both methods (0.57 and 0.48, respectively). Conclusions: Both nonlinear orthogonal projection and curvature-based retinal vessel segmentation methods were found to be sensitive to variations in image brightness, resulting from iris shadowing associated with different angle of photographic incidence.

Measurement of macular fractal dimension using a computer-assisted program.

PURPOSE Macular diseases may be associated with an altered retinal vasculature. We describe and test new software for the measurement of retinal vascular fractal dimension to quantify the complexity of retinal vasculature at the macula (D mac) and to compare this with fractal dimension measured around the optic disc (D disc). METHODS A total of 342 macular-centered and optic disc-centered digital retinal photographs from 171 subjects was selected randomly from a population-based study. Retinal vascular fractional dimension (Df) was measured by two trained graders using a computer-assisted program (SIVA-FA, software version 1.0, National University of Singapore) on macula-centered (D mac) and optic disc-centered (D disc) photographs, to assess intergrader reliability. Measurements were repeated after two weeks to determine intragrader reliability. A separate 50 pairs of consecutively repeated images were selected and measured using SIVA-FA to assess intrasession reliability. Reliability analyses were conducted using intraclass correlation coefficients (ICC), and multiple linear regression analyses were performed to compare factors associated with D mac and D disc measurements. RESULTS The mean (SD) D mac and D disc values were 1.453 (0.060) and 1.484 (0.043), respectively, and were highly correlated (r = 0.70, P < 0.001). Intragrader, intergrader, and intrasession reliability for both Df measures was high (ICCs ranging from 0.88-0.99). In multiple regression analyses, age (both β = -0.03, P < 0.001) and hypertension (β = -0.02, P = 0.011; β = -0.02, P = 0.021, respectively) were independently associated with D mac and D disc. CONCLUSIONS The complexity of the retinal vasculature in the macula can be measured reliably and may be a useful tool to study parafoveal vascular networks in macula diseases, such as diabetic maculopathy.

Distributed Coordination Guidance in Multi-agent Reinforcement Learning

In this paper we present a distributed reinforcement learning system that leverages on expert coordination knowledge to improve learning in multi-agent problems. We focus on the scenario where agents can communicate with their neighbors but this communication structure and the number of agents may change over time. We express coordination knowledge as constraints to reduce the joint action space for exploration. We introduce an extra learning level to learn when to make use of these constraints. This extra level is decentralized among the agents, making it suitable for our communication restrictions. Experiment results on tactical real-time strategy and soccer games show that our system is effective in online learning as opposed to existing methods that use individual constraints on agents and coordinated action selection.

Detecting Aggregate Incongruities in XML

The problem of identifying deviating patterns in XML repositories has important applications in data cleaning, fraud detection, and stock market analysis. Current methods determine data discrepancies by assessing whether the data conforms to the expected distribution of its immediate neighborhood. This approach may miss interesting deviations involving aggregated information. For example, the average number of transactions of a particular bank account may be exceptionally high as compared to other accounts with similar profiles. Such incongruity could only be revealed through aggregating appropriate data and analyzing the aggregated results in the associated neighborhood. This neighborhood is implicitly encapsulated in the XML structure. In addition, the hierarchical nature of the XML structure reflects the different levels of abstractions in the real world. This work presents a framework that detects incongruities in aggregate information. It utilizes the inherent characteristics of the XML structure to systematically aggregate leaf-level data and propagate the aggregated information up the hierarchy. The aggregated information is analyzed using a novel method by first clustering similar data, then, assuming a statistical distribution and identifying aggregate incongruity within the clusters. Experiments results indicate that the proposed approach is effective in detecting interesting discrepancies in a real world bank data set.

DeepDetect: An Extensible System for Detecting Attribute Outliers & Duplicates in XML

XML, the eXtensible Markup Language, is fast evolving into the new standard for data representation and exchange on the WWW. This has resulted in a growing number of data cleaning techniques to locate “dirty” data (artifacts). In this paper, we present DeepDetect – an extensible system that detects attribute outliers and duplicates in XML documents. Attribute outlier detection finds objects that contain deviating values with respect to a relevant group of objects. This entails utilizing the correlation among element values in a given XML document. Duplicate detection in XML requires the identification of subtrees that correspond to real world objects. Our system architecture enables sharing of common operations that prepare XML data for the various artifact detection techniques. DeepDetect also provides an intuitive visual interface for the user to specify various parameters for preprocessing and detection, as well as to view results.