Ole Vilhelm Larsen

Screening for diabetic retinopathy using computer based image analysis and statistical classification

Diabetic retinopathy is one of the most common causes of blindness in Europe. However, efficient therapies do exist. An accurate and early diagnosis and correct application of treatment can prevent blindness in more than 50% of all cases. Digital imaging is becoming available as a means of screening for diabetic retinopathy. As well as providing a high quality permanent record of the retinal appearance, which can be used for monitoring of progression or response to treatment, and which can be reviewed by an ophthalmologist, digital images have the potential to be processed by automatic analysis systems. We have described the preliminary development of a tool to provide automatic analysis of digital images taken as part of routine monitoring of diabetic retinopathy in our clinic. Various statistical classifiers, a Bayesian, a Mahalanobis, and a KNN classifier were tested. The system was tested on 134 retinal images. The Mahalanobis classifier had the best results: microaneurysms, haemorrhages, exudates, and cotton wool spots were detected with a sensitivity of 69, 83, 99, and 80%, respectively.

Modelling of interaction between a spatula and a human brain

This paper describes a method for surgery simulation, or more specifically a learning system of how to use a brain spatula. Improper use of brain spatulas can lead to brain tissue lesions such as tearing of brain tissue and ischemia. The idea is to provide surgeons with a tool which can teach them the correlation between deformation and applied force. The system includes a Finite Element based model of the brain in a Virtual Reality setup with haptic feedback. The physical model links the shape of the deformable model with the associated force. The interaction between the spatula and the brain model is handled by a collision response method which aims at smoothing the discrete haptic feedback. The experimental results are promising even though the used force feedback device is somewhat constraining the realism.

Guidelines for Choosing Optimal OParameters of Elasticity for Snakes

This paper proposes a guidance in the process of choosing and using the parameters of elasticity of a snake in order to obtain a precise segmentation. A new two step procedure is defined based on upper and lower bounds on the parameters. Formulas, by which these bounds can be calculated for real images where parts of the contour may be missing, are presented. Experiments on segmentation of bone structures in X-ray images have verified the usefulness of the new procedure.

Using Region-of-Interest Based Finite Element Modeling for Brain-Surgery Simulation

Brain surgery simulation requires a mathematical model of the geometric and elastic properties of the entire brain. To allow for realtime manipulation of the model it is necessary to differentiate the level of accuracy between different subparts of the brain model. A Finite Element Model (FEM) of the brain is presented capable of differentiating the spatial and temporal accuracy in different parts of the model. In a user defined region-of-interest around the surgical target point a dynamic FEM model is used to give high accuracy. The remaining parts of the brain is modelled by a static FEM model having less accuracy. The two models are integrated into one model for the entire brain using Condensation. In the context of our early version of a brain surgery simulator we have tested the condensed model versus a full dynamic model of the brain. Promising results concerning spatial error and execution time are shown.

Knowledge-based extraction of cerebral vasculature from anatomical MRI

A vessel extraction approach is presented that permits visualization of the cerebral vasculature in 3D from anatomical proton density (PD) weighted magnetic resonance imaging (MRI) volumes. The approach presented utilizes general knowledge about the shape and size of the cerebral vasculature and is divided into multi-scale vessel enhancement filtering, centre-line extraction, and surface modeling. To improve the discrimination between blood vessels and other tissue a multi-scale filtering method that enhances tubular structures is used as a pre-processing step. Centre-line extraction is applied to roughly estimate the centre-line of the vasculature involving both segmentation and skeletonization. The centre-line is used to initialize an active contour modeling process where cylinders are used to model the 3D surface of the blood vessels. The accuracy and robustness of the vessel extraction approach have been demonstrated on both simulated and real data (1mm3 voxels). On simulated data, the mean error of the estimated radii was found to be less than 0.4mm. On real data, the vasculature was successfully extracted from 20 MRI data sets using the same input parameters. An expert found the extracted vessel surfaces to coincide with the vessel walls in the data. Results from CTA data indicate that the approach will work successfully with other imaging modalities as well.

Using Weighted Minutiae for Fingerprint Identification

This paper describes how a minutiae reliability measure can be defined and integrated in an automatic fingerprint identification system. The reliability measure denotes the quality of the minutiae and is used to weight the minutiae in the matching of two fingerprints. To strengthen the matching even further, relational information is used along with attribute information for describing a fingerprint. Experimental results strongly suggest that the use of weighted minutiae can improve the matching performance significantly.

Bounds on the Optimal Elasticity Parameters for a Snake

This paper develops a formalism by which an estimate for the upper and lower bounds for the elasticity parameters for a snake can be obtained. Objects different in size and shape give rise to different bounds. The bounds can be obtained based on an analysis of the shape of the object of interest. Experiments on synthetic images show a good correlation between the estimated behaviour of the snake and the one actually observed. Experiments on real X-ray images show that the parameters for optimal segmentation lie within the estimated bounds.

A multi-resolution retinal vessel tracker based on directional smoothing

To support ophthalmologists in their routine and enable the quantitative assessment of vascular changes in color fundus photographs a multi-resolution approach was developed which segments the vessel tree efficiently and precisely in digital images of the retina. The algorithm starts at seed points, found in a preprocessing step and then follows the vessel, iteratively adjusting the direction of the search, and finding the center line of the vessels. As an addition, vessel branches and crossings are detected and stored in detailed lists. Every iteration of the Directional Smoothing Based (DSB) tracking process starts at a given point in the middle of a vessel. First rectangular windows for several directions in a neighborhood of this point are smoothed in the assumed direction of the vessel. The window, that results in the best contrast is then said to have the true direction of the vessel. The center point is moved into that direction 1/8th of the vessel width, and the algorithm continues with the next iteration. The vessel branch and crossing detection uses a list with unique vessel segment IDs and branch point IDs. During the tracking, when another vessel is crossed, the tracking is stopped. The newly traced vessel segment is stored in the vessel segment list, and the vessel, that had been traced before is broken up at the crossing- or branch point, and is stored as two different vessel segments. This approach has several advantages: - With directional smoothing, noise is eliminated, while the edges of the vessels are kept. - DSB works on high resolution images (3000 x 2000 pixel) as well as on low-resolution images (900 x 600 pixel), because a large area of the vessel is used to find the vessel direction - For the detection of venous beading the vessel width is measured for every step of the traced vessel. - With the lists of branch- and crossing points, we get a network of connected vessel segments, that can be used for further processing the retinal vessel tree.

Applying Voting to Segmentation of MR Images

The performance of applying voting to MR segmentation is investigated. Three different segmentation methods (fuzzy c-means, Bayes, and k-nearest neighbour) are used as input to the voting algorithm. Using human expert segmented images as a reference an error rate of 7.1% is obtained when applying voting. When comparing to the other methods it is seen that the results of applying the voting algorithm are slightly improved in terms of the error rate, minimum and maximum error.

Region-of-interest based finite element modelling of the brain-an approach to brain surgery simulation

One of the major problems in brain surgery simulation is to achieve both real-time performance and physical realism without compromising the necessary freedom to perform topological changes like cutting. In an attempt to solve the problem we present a new method for differentiating the spatial and temporal accuracy for the brain tissue models. Finite element methods are used for the modelling with a dynamic model applied to areas which require high accuracy and a static model for the remaining areas. The two models are integrated into one for the entire brain using condensation. The dynamic model is specified as a user-defined region-of-interest. The experiments carried out have shown promising results concerning spatial error and execution time.