In this dissertation, I investigated computing algorithms for automated retinal blood vessel detection. Changes in blood vessel structures are important indicators of many diseases such as diabetes, hypertension, etc. Blood vessel is also very useful in tracking of disease progression, and for biometric authentication. In this dissertation, I proposed two algorithms to detect blood vessel maps in retina. The first algorithm is based on integration of a Gaussian tracing scheme and a Gabor-variance filter. This algorithm traces the large blood vessel in retinal images enhanced with adaptive histogram equalization. Small vessels are traced on further enhanced images by a Gabor-variance filter. The second algorithm is called a radial contrast transform (RCT) algorithm, which converts the intensity information in spatial domain to a high dimensional radial contrast domain. Different feature descriptors are designed to improve the speed, sensitivity, and expandability of the vessel detection system. Performances comparison of the two algorithms with those in the literature shows favorable and robust results. Furthermore, a new performance measure based on central line of blood vessels is proposed as an alternative to more reliable assessment of detection schemes for small vessels, because the significant variations at the edges of small vessels need not be considered. The proposed algorithms were successfully tested in the field for early diabetic retinopathy (DR) screening. A highly modular code library to take advantage of the parallel processing power of multi-core computer architecture was tested in a clinical trial. Performance results showed that our scheme can achieve similar or even better performance than human expert readers for detection of micro-aneurysms on difficult images.

Microcirculation has become a key factor for the study and assessment of tissue perfusion and oxygenation. Detection and assessment of the microvasculature using videomicroscopy from the oral mucosa provides a metric on the density of blood vessels in each single frame. Information pertaining to the density of these microvessels within a field of view can be used to quantitatively monitor and assess the changes occurring in tissue oxygenation and perfusion over time. Automated analysis of this information can be used for real-time diagnostic and therapeutic planning of a number of clinical applications including resuscitation. The objective of this study is to design an automated image processing system to segment microvessels, estimate the density of blood vessels in video recordings, and identify the distribution of blood flow. The proposed algorithm consists of two main stages: video processing and image segmentation. The first step of video processing is stabilization. In the video stabilization step, block matching is applied to the video frames. Similarity is measured by cross-correlation coefficients. The main technique used in the segmentation step is multi-thresholding and pixel verification based on calculated geometric and contrast parameters. Segmentation results and differences of video frames are then used to identify the capillaries with blood flow. After categorizing blood vessels as active or passive, according to the amount of blood flow, quantitative measures identifying microcirculation are calculated. The algorithm is applied to the videos obtained using Microscan Side-stream Dark Field (SDF) imaging technique captured from healthy and critically ill humans/animals. Segmentation results were compared and validated using a blind detailed inspection by experts who used a commercial semi-automated image analysis software program, AVA (Automated Vascular Analysis). The algorithm was found to extract approximately 97% of functionally active capillaries and blood vessels in every frame. The aim of this study is to eliminate the human interaction, increase accuracy and reduce the computation time. The proposed method is an entirely automated process that can perform stabilization, pre-processing, segmentation, and microvessel identification without human intervention. The method may allow for assessment of microcirculatory abnormalities occurring in critically ill and injured patients including close to real-time determination of the adequacy of resuscitation.

A system for vessel characterization aimed at detecting deep vein thrombosis (DVT) in the lower limbs has been developed and evaluated using ultrasound image processing, location and force sensors measurements, blood flow information and a protocol based on the current clinical standard, compression ultrasound. The goal is to provide an objective and repeatable system to measure DVT in a rapid and standardized manner, as this has been suggested in the literature as an approach to improve overall detection of the disease. The system uses a spatial Kalman filter-based algorithm with an elliptical model in the measurement equation to detect vessel contours in transverse ultrasound images and estimate ellipse parameters, and temporal constant velocity Kalman filters for tracking vessel location in real-time. The vessel characterization also comprises building a 3-D vessel model and performing compression and blood flow assessments to calculate measures that indicate the possibility of DVT in a vessel. A user interface designed for assessing a vessel for DVT was also developed. The system and components were implemented and tested in simulations, laboratory settings, and clinical settings. Contour detection results are good, with mean and rms errors ranging from 1.47-3.64 and 3.69-9.67 pixels, respectively, in simulated and patient images, and parameter estimation errors of 5%. Experiments showed errors of 3-5 pixels for the tracking approaches. The measures for DVT were evaluated, independently and integrated in the system. The complete system was evaluated, with sensitivity of 67-100% and specificity of 50-89.5%. System learnability and memorability were evaluated in a separate user study, with good results. Contributions include a segmentation approach using a full parameter ellipse model in an extended Kalman filter, incorporating multiple measurements, an alternate sampling method for faster parameter convergence and application-specific initialization, and a tracking approach that includes a sub-sampled sum of absolutes similarity calculation and a method to detect vessel bifurcations using flow data. Further contributions include an integrated system for DVT detection that can combine ultrasound B-mode, colour flow and elastography images for vessel characterization, a system interface design focusing on usability that was evaluated with medical professionals, and system evaluations through multiple patient studies.

A system for vessel characterization aimed at detecting deep vein thrombosis (DVT) in the lower limbs has been developed and evaluated using ultrasound image processing, location and force sensors measurements, blood flow information and a protocol based on the current clinical standard, compression ultrasound. The goal is to provide an objective and repeatable system to measure DVT in a rapid and standardized manner, as this has been suggested in the literature as an approach to improve overall detection of the disease. The system uses a spatial Kalman filter-based algorithm with an elliptical model in the measurement equation to detect vessel contours in transverse ultrasound images and estimate ellipse parameters, and temporal constant velocity Kalman filters for tracking vessel location in real-time. The vessel characterization also comprises building a 3-D vessel model and performing compression and blood flow assessments to calculate measures that indicate the possibility of DVT in a vessel. A user interface designed for assessing a vessel for DVT was also developed. The system and components were implemented and tested in simulations, laboratory settings, and clinical settings. Contour detection results are good, with mean and rms errors ranging from 1.47-3.64 and 3.69-9.67 pixels, respectively, in simulated and patient images, and parameter estimation errors of 5%. Experiments showed errors of 3-5 pixels for the tracking approaches. The measures for DVT were evaluated, independently and integrated in the system. The complete system was evaluated, with sensitivity of 67-100% and specificity of 50-89.5%. System learnability and memorability were evaluated in a separate user study, with good results. Contributions include a segmentation approach using a full parameter ellipse model in an extended Kalman filter, incorporating multiple measurements, an alternate sampling method for faster parameter convergence and application-specific initialization, and a tracking approach that includes a sub-sampled sum of absolutes similarity calculation and a method to detect vessel bifurcations using flow data. Further contributions include an integrated system for DVT detection that can combine ultrasound B-mode, colour flow and elastography images for vessel characterization, a system interface design focusing on usability that was evaluated with medical professionals, and system evaluations through multiple patient studies.

A system for vessel characterization aimed at detecting deep vein thrombosis (DVT) in the lower limbs has been developed and evaluated using ultrasound image processing, location and force sensors measurements, blood flow information and a protocol based on the current clinical standard, compression ultrasound. The goal is to provide an objective and repeatable system to measure DVT in a rapid and standardized manner, as this has been suggested in the literature as an approach to improve overall detection of the disease. The system uses a spatial Kalman filter-based algorithm with an elliptical model in the measurement equation to detect vessel contours in transverse ultrasound images and estimate ellipse parameters, and temporal constant velocity Kalman filters for tracking vessel location in real-time. The vessel characterization also comprises building a 3-D vessel model and performing compression and blood flow assessments to calculate measures that indicate the possibility of DVT in a vessel. A user interface designed for assessing a vessel for DVT was also developed. The system and components were implemented and tested in simulations, laboratory settings, and clinical settings. Contour detection results are good, with mean and rms errors ranging from 1.47-3.64 and 3.69-9.67 pixels, respectively, in simulated and patient images, and parameter estimation errors of 5%. Experiments showed errors of 3-5 pixels for the tracking approaches. The measures for DVT were evaluated, independently and integrated in the system. The complete system was evaluated, with sensitivity of 67-100% and specificity of 50-89.5%. System learnability and memorability were evaluated in a separate user study, with good results. Contributions include a segmentation approach using a full parameter ellipse model in an extended Kalman filter, incorporating multiple measurements, an alternate sampling method for faster parameter convergence and application-specific initialization, and a tracking approach that includes a sub-sampled sum of absolutes similarity calculation and a method to detect vessel bifurcations using flow data. Further contributions include an integrated system for DVT detection that can combine ultrasound B-mode, colour flow and elastography images for vessel characterization, a system interface design focusing on usability that was evaluated with medical professionals, and system evaluations through multiple patient studies. / Applied Science, Faculty of / Electrical and Computer Engineering, Department of / Graduate

Ocular fundus image can provide information on pathological changes caused by local ocular diseases and early signs of certain systemic diseases. Automated analysis and interpretation of fundus images has become a necessary and important diagnostic procedure in ophthalmology. Among the features in ocular fundus image are the optic disc, fovea (central vision area), lesions, and retinal vessels. These features are useful in revealing the states of diseases in the form of measurable abnormalities such as length of diameter, change in color, and degree of tortuosity in the vessels. In addition, retinal vessels can also serve as landmarks for image-guided laser treatment of choroidal neovascularization. Thus, reliable methods for blood vessel detection that preserve various vessel measurements are needed. In this paper, we will examine the pathological issues in the analysis of retinal vessels in digital fundus images and give a survey of current image processing methods for extracting vessels in retinal images with a view to categorize them and highlight their differences and similarities. We have also implemented two major approaches using matched filter and mathematical morphology respectively and compared their performances. Some prospective research directions are identified. / Singapore-MIT Alliance (SMA)

Η κατασκευή συστημάτων τομογραφίας του ανθρώπινου ιστού τα οποία θα χρησιμοποιούν το υπέρυθρο φάσμα ακτινοβολίας αποτελεί σημαντική προοπτική για τη δημιουργία νέων ιατρικών διαγνωστικών μεθόδων. Ένα από τα σημαντικότερα προβλήματα που πρέπει να επιλυθούν είναι η μικρή διεισδυτική ικανότητα και ο υψηλός βαθμός απορρόφησης και σκέδασης που παραμορφώνει ισχυρά την ακτινοβολία που διαδίδεται μέσα από τον ανθρώπινο ιστό. Στα πλαίσια της διδακτορικής διατριβής, μελετήθηκε το πρόβλημα του εντοπισμού της θέσης των αγγείων σε ψηφιακές φωτογραφίες του ανθρώπινου δακτύλου που έχουν ληφθεί στο υπέρυθρο φάσμα. Για τον σκοπό αυτό αναπτύχθηκε μεγάλος αριθμός πρωτότυπων μεθόδων κανονικοποίησης της φωτεινότητας της εικόνας, μη-γραμμικής ενίσχυσης της αντίθεσης, αφαίρεσης των γραμμών δακτυλικών αποτυπωμάτων, εντοπισμού του προτύπου ή δικτύου αγγείων και βελτίωσης του προτύπου των αγγείων χρησιμοποιώντας μεθόδους μαθηματικής μορφολογίας. Συνοπτικά στην παρούσα διδακτορική διατριβή προτάθηκαν και εξετάσθηκαν διαφορετικές πρωτότυπες μέθοδοι και αλγόριθμοι με επίβλεψη ή χωρίς επίβλεψη για την εξαγωγή του προτύπου αγγείων από υπέρυθρες εικόνες του ανθρώπινου δακτύλου καθώς και διαφορετικές πρωτότυπες μέθοδοι και αλγόριθμοι χωρίς επίβλεψη για την εξαγωγή του δικτύου αγγείων από αμφιβληστροειδικές εικόνες του ανθρώπινου οφθαλμού. Επίσης, η ερευνητική προσπάθεια επικεντρώθηκε στην βελτίωση των εικόνων που λαμβάνονται από το προτεινόμενο σύστημα απόκτησης εικόνων, γεγονός το οποίο οδήγησε στην ανάπτυξη πρωτότυπων μεθόδων προ-επεξεργασίας και τη μετέπειτα βελτίωση των αρχικών αποτελεσμάτων κατάτμησης που προκύπτουν από την εφαρμογή των μεθόδων ή αλγορίθμων κατάτμησης προτύπου αγγείων, γεγονός το οποίο οδήγησε στην ανάπτυξη πρωτότυπων μεθόδων μετά-επεξεργασίας. / The construction of tomographic systems of human tissue which use the infrared spectrum of radiation constitutes an important capability of making new medical diagnostic methods. One of the most crucial problems which must be resolved is the low penetrating ability and the high degree of absorption and scattering which strongly distort the radiation that pass through the human tissue. In this thesis, the problem of the extraction of finger vein pattern from infrared images of finger and the similar problem of retinal vessel tree segmentation were studied. Moreover, the problem of shading and non-uniform illumination correction was also studied in images which suffer from the above problems either due to imperfect set-up of the image acquisition system or due to the interaction between objects and illumination on the scene. In this thesis, existing algorithms were improved and novel algorithms were developed. Both vein pattern extraction algorithms and shading and non-uniform illumination correction algorithms were proposed. The proposed methods include novel preprocessing modules for intensity normalization, elimination of fingerprint lines, non linear contrast enhancement using spatial information, and shading and non uniform illumination correction. The vein pattern extraction was performed using ten novel methods that use structural classification methods, spatial derivatives information and fuzzy set theory. The effectiveness of the proposed methods and algorithms was evaluated both on real and artificial images distorted by different types of noise and different signal to noise ratios. The majority of the methods present satisfactory accuracy on the detection of vein network, something happens due to the successful collaboration between the preprocessing methods and the vein pattern extraction methods. In addition, the problem of improving the vein network extraction accuracy was successfully handled using advanced postprocessing methods based on binary mathematical morphology. Finally, in this thesis two novel methods for retinal vessel segmentation were proposed and evaluated. They also compared with the most important methods have already been presented in the literature and one of them achieved the best experimental results from all the unsupervised methods evaluated in the publicly available DRIVE database.