Showing papers by "Dorin Comaniciu published in 2006"

Total variation models for variable lighting face recognition

Abstract: In this paper, we present the logarithmic total variation (LTV) model for face recognition under varying illumination, including natural lighting conditions, where we rarely know the strength, direction, or number of light sources. The proposed LTV model has the ability to factorize a single face image and obtain the illumination invariant facial structure, which is then used for face recognition. Our model is inspired by the SQI model but has better edge-preserving ability and simpler parameter selection. The merit of this model is that neither does it require any lighting assumption nor does it need any training. The LTV model reaches very high recognition rates in the tests using both Yale and CMU PIE face databases as well as a face database containing 765 subjects under outdoor lighting conditions

Probabilistic 3D Polyp Detection in CT Images: The Role of Sample Alignment

Abstract: Automatic polyp detection is an increasingly important task in medical imaging with virtual colonoscopy [15] being widely used. In this paper, we present a 3D object detection algorithm and show its application on polyp detection from CT images. We make the following contributions: (1) The system adopts Probabilistic Boosting Tree (PBT) to probabilistically detect polyps. Integral volume and 3D Haar filters are introduced to achieve fast feature computation. (2) We give an explicit convergence rate analysis for the AdaBoost algorithm [2] and prove that the error at each step \in t+1. is tightly bounded by the previous error \in t. (3) For a 3D polyp template, a generative model is defined. Given the bound and convergence analysis, we analyze the role of "sample alignment" in the template design and devise a robust and efficient algorithm for polyp detection. The overall system has been tested on 150 volumes and the results obtained are very encouraging.

Reliable Detection of Overtaking Vehicles Using Robust Information Fusion

Abstract: Early detection of overtaking vehicles is an important task for vision-based driver assistance systems. Techniques utilizing image motion are likely to suffer from spurious image structures caused by shadows and illumination changes, let alone the aperture problem. To achieve reliable detection of overtaking vehicles, the authors have developed a robust detection method, which integrates dynamic scene modeling, hypothesis testing, and robust information fusion. A robust fusion algorithm, based on variable bandwidth density fusion and multiscale mean shift, is introduced to obtain reliable motion estimation against various image noise. To further reduce detection error, the authors model the dynamics of road scenes and exploit useful constraints induced by the temporal coherence in vehicle overtaking. The proposed solution is integrated into a monocular vision system onboard for obstacle detection. Test results have shown superior performance achieved by the new method

System and method for integrating heterogeneous biomedical information

Abstract: A system and method for using heterogeneous data from multiple healthcare information sources in a medical decision support system is disclosed. Each healthcare information system stores medical data using a different local schema. The medical decision support system provides responses to user queries. A query is received from a user that is generated in a standardized global schema. The query includes information from medical ontologies. Database queries are generated from the user queries that use the medical ontologies to generate constraints in the queries. The medical ontologies are also used to infer database queries. The generated query is translated into multiple queries for the multiple healthcare systems wherein each query is in the local schema of the healthcare information system that is being queried. Each database query is transmitted to one of the healthcare information systems based on the local schema of the particular query. Data is collected from each of the queried healthcare information system and analyzed. A query response is formulated for the user

Image-Based Multiclass Boosting and Echocardiographic View Classification

Abstract: We tackle the problem of automatically classifying cardiac view for an echocardiographic sequence as a multiclass object detection. As a solution, we present an imagebased multiclass boosting procedure. In contrast with conventional approaches for multiple object detection that train multiple binary classifiers, one per object, we learn only one multiclass classifier using the LogitBoosting algorithm. To utilize the fact that, in the midst of boosting, one class is fully separated from the remaining classes, we propose to learn a tree structure that focuses on the remaining classes to improve learning efficiency. Further, we accommodate the large number of background images using a cascade of boosted multiclass classifiers, which is able to simultaneously detect and classify multiple objects while rejecting the background class quickly. Our experiments on echocardiographic view classification demonstrate promising performances of image-based multiclass boosting.

Medical diagnostic imaging optimization based on anatomy recognition

Abstract: Anatomical information is identified from a medical image and/or used for controlling a medical diagnostic imaging system, such as an ultrasound system. To identify anatomical information from a medical image, a processor applies a multi-class classifier. The anatomical information is used to set an imaging parameter of the medical imaging system. The setting or identification may be used in combination or separately.

Method for database-guided simultaneous multi-slice object detection in three dimensional volumetric data

Abstract: The present invention is directed to a method for automatic detection and segmentation of a target anatomical structure in received three dimensional (3D) volumetric medical images using a database of a set of volumetric images with expertly delineated anatomical structures. A 3D anatomical structure detection and segmentation module is trained offline by learning anatomical structure appearance using the set of expertly delineated anatomical structures. A received volumetric image for the anatomical structure of interest is searched online using the offline learned 3D anatomical structure detection and segmentation module.

System and method for fetal biometric measurements from ultrasound data and fusion of same for estimation of fetal gestational age

Abstract: A method for segmenting and measuring anatomical structures in fetal ultrasound images includes the steps of providing a digitized ultrasound image of a fetus comprising a plurality of intensities corresponding to a domain of points on a 3-dimensional grid, providing a plurality of classifiers trained to detect anatomical structures in said image of said fetus, and segmenting and measuring an anatomical structure using said image classifiers by applying said elliptical contour classifiers to said fetal ultrasound image, wherein a plurality of 2-dimensional contours characterizing said anatomical structure are detected. The anatomical structure measurement can be combined with measurement of another anatomical structure to estimate gestational age of the fetus.

Health-e-Child: An Integrated Biomedical Platform for Grid-Based Pediatrics

Abstract: There is a compelling demand for the integration and exploitation of heterogeneous biomedical information for improved clinical practice, medical research, and personalised healthcare across the EU. The Health-e-Child project aims at developing an integrated healthcare platform for European Paediatrics, providing seamless integration of traditional and emerging sources of biomedical information. The long-term goal of the project is to provide uninhibited access to universal biomedical knowledge repositories for personalised and preventive healthcare, large-scale information-based biomedical research and training, and informed policy making. The project focus will be on individualised disease prevention, screening, early diagnosis, therapy and follow-up of paediatric heart diseases, inflammatory diseases, and brain tumours. The project will build a Gridenabled European network of leading clinical centres that will share and annotate biomedical data, validate systems clinically, and diffuse clinical excellence across Europe by setting up new technologies, clinical workflows, and standards. This paper outlines the design approach being adopted in Health-e-Child to enable the delivery of an integrated biomedical information platform.

BoostMotion: Boosting a Discriminative Similarity Function for Motion Estimation

Abstract: Motion estimation for applications where appearance undergoes complex changes is challenging due to lack of an appropriate similarity function. In this paper, we propose to learn a discriminative similarity function based on an annotated database that exemplifies the appearance variations. We invoke the LogitBoost algorithm to selectively combine weak learners into one strong similarity function. The weak learners based on local rectangle features are constructed as nonparametric 2D piecewise constant functions, using the feature responses from both images, to strengthen the modeling power and accommodate fast evaluation. Because the negatives possess a location parameter measuring their closeness to the positives, we present a locationsensitive cascade training procedure, which bootstraps negatives for later stages of the cascade from the regions closer to the positives. This allows viewing a large number of negatives and steering the training process to yield lower training and test errors. In experiments of estimating the motion for the endocardial wall of the left ventricle in echocardiography, we compare the learned similarity function with conventional ones and obtain improved performances. We also contrast the proposed method with a learning-based detection algorithm to demonstrate the importance of temporal information in motion estimation. Finally, we insert the learned similarity function into a simple contour tracking algorithm and find that it reduces drifting.

Method and apparatus for detecting blood vessel boundaries using multi-scale mean-shift ray propagation

Abstract: Disclosed is a method and system for detecting a boundary of a vessel in an image. Edges in the image are detected. Edge detection is based on the change in intensity over some distance while varying the scale of the distance. A set of edges is then selected from the detected edges. An initial vessel boundary is determined based on the selected set, and a shape descriptor (e.g., one or more elliptical shape descriptors) is applied to the initial vessel boundary to determine a final vessel boundary.

Database-guided simultaneous multi-slice 3d segmentation for volumetric data

Abstract: Automatic delineation of anatomical structures in 3-D volumetric data is a challenging task due to the complexity of the object appearance as well as the quantity of information to be processed. This makes it increasingly difficult to encode prior knowledge about the object segmentation in a traditional formulation as a perceptual grouping task. We introduce a fast shape segmentation method for 3-D volumetric data by extending the 2-D database-guided segmentation paradigm which directly exploits expert annotations of the interest object in large medical databases. Rather than dealing with 3-D data directly, we take advantage of the observation that the information about position and appearance of a 3-D shape can be characterized by a set of 2-D slices. Cutting these multiple slices simultaneously from the 3-D shape allows us to represent and process 3-D data as efficiently as 2-D images while keeping most of the information about the 3-D shape. To cut slices consistently for all shapes, an iterative 3-D non-rigid shape alignment method is also proposed for building local coordinates for each shape. Features from all the slices are jointly used to learn to discriminate between the object appearance and background and to learn the association between appearance and shape. The resulting procedure is able to perform shape segmentation in only a few seconds. Extensive experiments on cardiac ultrasound images demonstrate the algorithm's accuracy and robustness in the presence of large amounts of noise.

A learning based approach for 3d segmentation and colon detagging

Abstract: Foreground and background segmentation is a typical problem in computer vision and medical imaging. In this paper, we propose a new learning based approach for 3D segmentation, and we show its application on colon detagging. In many problems in vision, both the foreground and the background observe large intra-class variation and inter-class similarity. This makes the task of modeling and segregation of the foreground and the background very hard. The framework presented in this paper has the following key components: (1) We adopt probabilistic boosting tree [9] for learning discriminative models for the appearance of complex foreground and background. The discriminative model ratio is proved to be a pseudo-likelihood ratio modeling the appearances. (2) Integral volume and a set of 3D Haar filters are used to achieve efficient computation. (3) We devise a 3D topology representation, grid-line, to perform fast boundary evolution. The proposed algorithm has been tested on over 100 volumes of size 500 × 512 × 512 at the speed of 2 ~ 3 minutes per volume. The results obtained are encouraging.

Method and system for generating and validating clinical reports with built-in automated measurement and decision support

Abstract: A method and system for generation and validation of clinical reports with built-in automated measurement and decision support is provided. XML templates may be used for ensuring report data quality. The template data may include static, dynamic, calculated data, and others. The validation method may be based on formal logical constraint specifications. The constraint descriptions may include data types, cardinality, order, co-occurrence, Boolean logic, read-only data, regular expression patterns, etc. The method can be used to validate collected data on-the-fly based on constraint specifications without human interaction.

Health-e-child: an integrated biomedical platform for grid-based paediatric applications.

Abstract: There is a compelling demand for the integration and exploitation of heterogeneous biomedical information for improved clinical practice, medical research, and personalised healthcare across the EU. The Health-e-Child project aims at developing an integrated healthcare platform for European Paediatrics, providing seamless integration of traditional and emerging sources of biomedical information. The long-term goal of the project is to provide uninhibited access to universal biomedical knowledge repositories for personalised and preventive healthcare, large-scale information-based biomedical research and training, and informed policy making. The project focus will be on individualized disease prevention, screening, early diagnosis, therapy and follow-up of paediatric heart diseases, inflammatory diseases, and brain tumours. The project will build a Grid-enabled European network of leading clinical centres that will share and annotate biomedical data, validate systems clinically, and diffuse clinical excellence across Europe by setting up new technologies, clinical workflows, and standards. This paper outlines the design approach being adopted in Health-e-Child to enable the delivery of an integrated biomedical information platform.

Example Based Non-rigid Shape Detection

Abstract: Since it is hard to handcraft the prior knowledge in a shape detection framework, machine learning methods are preferred to exploit the expert annotation of the target shape in a database. In the previous approaches [1, 2], an optimal similarity transformation is exhaustively searched for to maximize the response of a trained classification model. At best, these approaches only give a rough estimate of the position of a non-rigid shape. In this paper, we propose a novel machine learning based approach to achieve a refined shape detection result. We train a model that has the largest response on a reference shape and a smaller response on other shapes. During shape detection, we search for an optimal non-rigid deformation to maximize the response of the trained model on the deformed image block. Since exhaustive searching is inapplicable for a non-rigid deformation space with a high dimension, currently, example based searching is used instead. Experiments on two applications, left ventricle endocardial border detection and facial feature detection, demonstrate the robustness of our approach. It outperforms the well-known ASM and AAM approaches on challenging samples.

System and method for semantic indexing and navigation of volumetric images

Abstract: A method for navigating digital medical images includes providing a digitized patient medical image of a structure of interest in a patient, using a pre-trained classifier to segment the structure of interest from the image, creating links from the structure of interest to a corresponding structure in an anatomical atlas, receiving a query to view the structure of interest, parsing the query to identify one or more keywords from noun phrases in the query, mapping a keyword to a corresponding structure in the anatomical atlas, wherein the anatomical atlas structure is associated with a link to the corresponding structure in the patient image, and following the link to display said patient structure of interest.

System and method for using learned discriminative models to segment three dimensional colon image data

Abstract: A system and method for using learned discriminative models to segment a border of an anatomical structure in a three dimensional (3D) image is disclosed. A discriminative probability model is computed for each voxel in the 3D image. Thresholding is performed on each discriminative probability model. One or more two dimensional (2D) slices of the thresholded 3D image along X-Y planes are obtained. Seed regions are selected in the 2D slices. Morphological region growing is performed on the selected seed regions. An initial 3D segmentation is obtained. Boundary evolution is performed on the initial 3D segmentation. The segmented anatomical structure is removed. in the original 3D image.

System and method for learning relative distance in a shape space using image based features

Abstract: A system and method for identifying a shape of an anatomical structure in an input image is disclosed. An input image is received and warped using a set of warping templates resulting in a set of warped images. An integral image is calculated for each warped image. Selected features are extracted based on the integral image. A boosted feature score is calculated for the combined selected features for each warped image. The warped images are ranked based on the boosted feature scores. A predetermined number of warped images are selected that have the largest feature scores. Each selected warped image is associated with its corresponding warping template. The corresponding warping templates are associated with stored shape models. The shape of the input image is identified based on the weighted average of the shapes models.

Method for characterizing shape, appearance and motion of an object that is being tracked

Abstract: A method for generating Pairwise Active Appearance Models (PAAMs) that characterize shape, appearance and motion of an object and using the PAAM to track the motion of an object is disclosed. A plurality of video streams is received. Each video stream includes a series of image frames that depict an object in motion. Each video stream includes an index of identified motion phases that are associated with a motion cycle of the object. For each video stream, a shape of the object is represented by a shape vector. An appearance of an object is represented by an appearance vector. The shape and appearance vectors associated at two consecutive motion phases are concatenated. Paired data for the concatenated shape and appearance vectors is computed. Paired data is computed for each two consecutive motion phases in the motion cycle. A shape subspace is constructed based on the computed paired data. An appearance subspace is constructed based on the computed paired data. A joint subspace is constructed using a combination of the shape subspace and appearance subspace. A PAAM is generated using the joint subspace and the PAAM is stored in a database.

System and method for using a similarity function to perform appearance matching in image pairs

Abstract: The present invention is directed to a method for populating a database with a set of images of an anatomical structure. The database is used to perform appearance matching in image pairs of the anatomical structure. A set of image pairs of anatomical structures is received, where each image pair is annotated with a plurality of location-sensitive regions that identify a particular aspect of the anatomical structure. Weak learners are iteratively selected and an image patch is identified. A boosting process is used to identify a strong classifier based on responses to the weak learners applied to the identified image patch for each image pair. The responses comprise a feature response and a location response associated with the image patch. Positive and negative image pairs are generated. The positive and negative image pairs are used to learn a similarity function. The learned similarity function and iteratively selected weak learners are stored in the database.

Hierarchical part-based detection of 3d flexible tubes: application to CT colonoscopy

Abstract: In this paper, we present a learning-based method for the detection and segmentation of 3D free-form tubular structures, such as the rectal tubes in CT colonoscopy. This method can be used to reduce the false alarms introduced by rectal tubes in current polyp detection algorithms. The method is hierarchical, detecting parts of the tube in increasing order of complexity, from tube cross sections and tube segments to the whole flexible tube. To increase the speed of the algorithm, candidate parts are generated using a voting strategy. The detected tube segments are combined into a flexible tube using a dynamic programming algorithm. Testing the algorithm on 210 unseen datasets resulted in a tube detection rate of 94.7% and 0.12 false alarms per volume. The method can be easily retrained to detect and segment other tubular 3D structures.

Pairwise active appearance model and its application to echocardiography tracking

Abstract: We propose a pairwise active appearance model (PAAM) to characterize statistical regularities in shape, appearance, and motion presented by a target that undergoes a series of motion phases, such as the left ventricle in echocardiography. The PAAM depicts the transition in motion phase through a Markov chain and the transition in both shape and appearance through a conditional Gaussian distribution. We learn from a database the joint Gaussian distribution of the shapes and appearances belonging to two consecutive motion phases (i.e., a pair of motion phases), from which we analytically compute the conditional Gaussian distribution. We utilize the PAAM in tracking the left ventricle contour in echocardiography and obtain improved tracking results in terms of localization accuracy when compared with expert-specified contours.

Three-dimensional cardiac border delineation in medical imaging

Abstract: Three-dimensional cardiac border is delineated in medical imaging. A view is labeled, such as identifying a two-dimensional view as an apical four-chamber view. A three-dimensional border is detected as a function of the view label. For example, the view is associated from a plane through a volume and a known orientation relative to the heart. Labeling the view indicates the orientation of the heart in the scanned volume. By determining the orientation of the heart, border detection processes may be simplified or assisted.

Method for performing image based regression using boosting

Abstract: A method for performing image based regression using boosting to infer an entity that is associated with an image of an object is disclosed. A regression function for a plurality of images is learned in which for each image the associated entity is known. The learned regression function is used to predict an entity associated with an image in which the entity is not known.

System and method for automatic molecular diagnosis of ALS based on boosting classification

Abstract: A method for diagnosing Amyotrophic lateral sclerosis includes providing surface-enhanced laser desorption/ionisation mass spectrometric (SELDI-MS) data of a plurality of proteins, said data obtained from a patient and comprising a plurality of peak values, and analysing said peak values with an alternating decision tree comprising a set of tests of said data peaks values and associated prediction values, wherein said data is predictive of depression if a sum of the prediction values of said tree is greater than 10

Simultaneous Registration and Modeling of Deformable Shapes

Abstract: Many natural objects vary the shapes as linear combinations of certain bases. The measurement of such deformable shapes is coupling of rigid similarity transformations between the objects and the measuring systems and non-rigid deformations controlled by the linear bases. Thus registration and modeling of deformable shapes are coupled problems, where registration is to compute the rigid transformations and modeling is to construct the linear bases. The previous methods [3, 2] separate the solution into two steps. The first step registers the measurements regarding the shapes as rigid and the deformations as random noise. The second step constructs the linear model using the registered shapes. Since the deformable shapes do not vary randomly but are constrained by the underlying model, such separate steps result in registration biased by nonrigid deformations and shape models involving improper rigid transformations. We for the first time present this bias problem and formulate that, the coupled registration and modeling problems are essentially a single factorization problem and thus require a simultaneous solution. We then propose the Direct Factorization method that extends a structure from motion method [16]. It yields a linear closedform solution that simultaneously registers the deformable shapes at arbitrary dimensions (2D \to 2D, 3D \to 3D, . . .) and constructs the linear bases. The accuracy and robustness of the proposed approach are demonstrated quantitatively on synthetic data and qualitatively on real shapes.

Joint classification and subtype discovery in tumor diagnosis by gene expression profiling

Abstract: A program storage device is provided readable by machine, tangibly embodying a program of instructions executable by the machine to perform method steps for classification of biological tissue by gene expression profiling. The method steps include providing a training set of gene expression profiles of known tissue samples, providing a first-layer strong classifier of the known tissue samples by combining weak classifiers using boosting, creating two sample sets based on the first classifier, populating the two sample sets with a next-layer of classifiers based on a previous-layer classifier, organizing the classifiers in a tree data structure, and outputting the tree data structure as a probabilistic boosting tree classifier for tissue sample classification and disease subtype discovery. A multi-class diagnosis problem is transformed to a two-class diagnosis process by finding an optimal feature and dividing the multi-class problem into two-classes.

Method and apparatus for detecting vessel boundaries

Abstract: Disclosed is a method and system for detecting a boundary of a vessel in an image. Edges in the image are detected. Edge detection is based on the change in intensity over some distance while varying the scale of the distance. A set of edges is then selected from the detected edges. An initial vessel boundary is determined based on the selected set, and a shape descriptor (e.g., one or more elliptical shape descriptors) is applied to the initial vessel boundary to determine a final vessel boundary.

Exploiting Interactions among Polymorphisms Contributing to Complex Disease Traits with Boosted Generative Modeling

Abstract: Although there has been great success in identifying disease genes for simple, monogenic Mendelian traits, deciphering the genetic mechanisms involved in complex diseases remains challenging. One major approach is to identify configurations of interacting factors such as single nucleotide polymorphisms (SNPs) that confer susceptibility to disease. Traditional methods, such as the multiple dimensional reduction method and the combinatorial partitioning method, provide good tools to decipher such interactions amid a disease population with a single genetic cause. However, these traditional methods have not managed to resolve the issue of genetic heterogeneity, which is believed to be a very common phenomenon in complex diseases. There is rarely prior knowledge of the genetic heterogeneity of a disease, and traditional methods based on estimation over the entire population are unlikely to succeed in the presence of heterogeneity. We present a novel Boosted Generative Modeling (BGM) approach for structure-model the interactions leading to diseases in the context of genetic heterogeneity. Our BGM method bridges the ensemble and generative modeling approaches to genetic association studies under a case-control design. Generative modeling is employed to model the interaction network configuration and the causal relationships, while boosting is used to address the genetic heterogeneity problem. We perform our method on simulation data of complex diseases. The results indicate that our method is capable of modeling the structure of interaction networks among disease-susceptible loci and of addressing genetic heterogeneity issues where the traditional methods, such as multiple dimensional reduction method, fail to apply. Our BGM method provides an exploratory tool that identifies the variables (e.g., disease-susceptible loci) that are likely to correlate and contribute to the disease.