Skip to main content

Adrian Barbu

Florida State University, Statistics, Faculty Member

Followers

49

Following

8

Co-authors

4

Public Views

Adrian Barbu graduated from UCLA in 2005. Until 2007 he worked in Siemens Corporate Research on many challenging Medical Imaging problems. Since 2007 he has been in the Statistics Department at FSU. His research interests are Computer Vision, Medical Imaging and Machine Learning.
Supervisors: Song-Chun Zhu
Phone: +1 850-980-2516

less

Amrita Vishwa Vidyapeetham

Kansas State University

Anoop Chaturvedi

University of Allahabad

Armando Marques-Guedes

UNL - New University of Lisbon

Animal Health Research Institute

Prof. Tofazzal Islam

Bangabandhu Sheikh Mujibur Rahman Agricultural University

University of Technology Sydney

PALIMOTE JUSTICE

RIVERS STATE POLYTECHNIC

Syukri Muhammad Nur

Darma Persada University

Srinivasa Rao Gadde

The University of Dodoma, Tanzania

Interests

Uploads

Papers by Adrian Barbu

RENOIR-A Benchmark Dataset for Real Noise Reduction Evaluation

Hierarchical Object Parsing from Structured Noisy Point Clouds

Eprint Arxiv 1108 3605, Aug 1, 2011

Object parsing and segmentation from point clouds are challenging tasks because the relevant data... more Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as Active Shape and Active Appearance models lack the necessary flexibility for this task, while recent approaches such as the Recursive Compositional Models make model simplifications in order to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer, which is a deformation of a hidden PCA shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state of the art parsing errors on two standard datasets without using any intensity information.

Hierarchical Object Parsing from Structured Noisy Point Clouds

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013

Object parsing and segmentation from point clouds are challenging tasks because the relevant data... more Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as active shape and active appearance models (AAMs) lack the necessary flexibility for this task, while recent approaches such as the recursive compositional models make model simplifications to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer which is a deformation of a hidden principal component analysis (PCA) shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state-of-the-art parsing errors on two standard datasets without using any intensity information.

Learning mixtures of Bernoulli templates by two-round EM with performance guarantee

Electronic Journal of Statistics, 2014

showed that a two-round variant of the EM algorithm can learn mixture of Gaussian distributions w... more showed that a two-round variant of the EM algorithm can learn mixture of Gaussian distributions with near optimal precision with high probability if the Gaussian distributions are well separated and if the dimension is sufficiently high. In this paper, we generalize their theory to learning mixture of high-dimensional Bernoulli templates. Each template is a binary vector, and a template generates examples by randomly switching its binary components independently with a certain probability. In computer vision applications, a binary vector is a feature map of an image, where each binary component indicates whether a local feature or structure is present or absent within a certain cell of the image domain. A Bernoulli template can be considered as a statistical model for images of objects (or parts of objects) from the same category. We show that the two-round EM algorithm can learn mixture of Bernoulli templates with near optimal precision with high probability, if the Bernoulli templates are sufficiently different and if the number of features is sufficiently high. We illustrate the theoretical results by synthetic and real examples.

Rates for Inductive Learning of Compositional Models

Training an active random field for real-time image denoising

Abstract Many computer vision problems can be formulated in a Bayesian framework based on Markov ... more Abstract Many computer vision problems can be formulated in a Bayesian framework based on Markov random fields (MRF) or conditional random fields (CRF). Generally, the MRF/CRF model is learned independently of the inference algorithm that is used to obtain the final result. In this paper, we observe considerable gains in speed and accuracy by training the MRF/CRF model together with a fast and suboptimal inference algorithm.

SPADES and mixture models

Abstract This paper studies sparse density estimation via ℓ 1 penalization (SPADES). We focus on ... more Abstract This paper studies sparse density estimation via ℓ 1 penalization (SPADES). We focus on estimation in high-dimensional mixture models and nonparametric adaptive density estimation. We show, respectively, that SPADES can recover, with high probability, the unknown components of a mixture of probability densities and that it yields minimax adaptive density estimates. These results are based on a general sparsity oracle inequality that the SPADES estimates satisfy.

Automatic detection and segmentation of axillary lymph nodes

Lymph node detection and measurement is a difficult and important part of cancer treatment. In th... more Lymph node detection and measurement is a difficult and important part of cancer treatment. In this paper we present a robust and effective learning-based method for the automatic detection of solid lymph nodes from Computed Tomography data. The contributions of the paper are the following. First, it presents a learning based approach to lymph node detection based on Marginal Space Learning. Second, it presents an efficient MRF-based segmentation method for solid lymph nodes.

Accurate polyp segmentation for 3D CT colongraphy using multi-staged probabilistic binary learning and compositional model

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

Accurate and automatic colonic polyp segmentation and measurement in Computed Tomography (CT) has... more Accurate and automatic colonic polyp segmentation and measurement in Computed Tomography (CT) has significant importance for 3D polyp detection, classification, and more generally computer aided diagnosis of colon cancers. In this paper, we propose a three-staged probabilistic binary classification approach for automatically segmenting polyp voxels from their surrounding tissues in CT. Our system integrates low-, and mid-level information for discriminative learning under local polar coordinates which align on the 3D colon surface around detected polyp. More importantly, our supervised learning system has flexible modeling capacity, which offers a principled means of encoding semantic, clinical expert annotations of colonic polyp tissue identification and segmentation. The learning generality to unseen data is bounded by boosting and stacked generality . Extensive experimental results on polyp segmentation performance evaluation and robustness testing with disturbances (using both training data and unseen data) are provided to validate our presented approach. The reliability of polyp segmentation and measurement has been largely increased to 98.2% (ie. errors ≤ 3mm), compared with other state of art work of about 75% ∼ 80%.

Dimension reduction and variable selection in case control studies via regularized likelihood optimization

Abstract Dimension reduction and variable selection are performed routinely in case-control studi... more Abstract Dimension reduction and variable selection are performed routinely in case-control studies, but the literature on the theoretical aspects of the resulting estimates is scarce. We bring our contribution to this literature by studying estimators obtained via ℓ 1 penalized likelihood optimization. We show that the optimizers of the ℓ 1 penalized retrospective likelihood coincide with the optimizers of the ℓ 1 penalized prospective likelihood.

Simultaneous Detection and Registration for Ileo-Cecal Valve Detection in 3D CT Colonography

Lecture Notes in Computer Science, 2008

Object detection and recognition has achieved a significant progress in recent years. However rob... more Object detection and recognition has achieved a significant progress in recent years. However robust 3D object detection and segmentation in noisy 3D data volumes remains a challenging problem. Localizing an object generally requires its spatial configuration (i.e., pose, size) being aligned with the trained object model, while estimation of an object's spatial configuration is only valid at locations where the object appears. Detecting object while exhaustively searching its spatial parameters, is computationally prohibitive due to the high dimensionality of 3D search space. In this paper, we circumvent this computational complexity by proposing a novel framework capable of incrementally learning the object parameters (IPL) of location, pose and scale. This method is based on a sequence of binary encodings of the projected true positives from the original 3D object annotations (i.e., the projections of the global optima from the global space into the sections of subspaces). The training samples in each projected subspace are labeled as positive or negative, according their spatial registration distances towards annotations as ground-truth. Each encoding process can be considered as a general binary classification problem and is implemented using probabilistic boosting tree algorithm. We validate our approach with extensive experiments and performance evaluations for Ileo-Cecal Valve (ICV) detection in both clean and tagged 3D CT colonography scans. Our final ICV detection system also includes an optional prior learning procedure for IPL which further speeds up the detection.

Generalizing Swendsen-Wang to sampling arbitrary posterior probabilities

Abstract Many vision tasks can be formulated as graph partition problems that minimize energy fun... more Abstract Many vision tasks can be formulated as graph partition problems that minimize energy functions. For such problems, the Gibbs sampler provides a general solution but is very slow, while other methods, such as Ncut and graph cuts are computationally effective but only work for specific energy forms and are not generally applicable. In this paper, we present a new inference algorithm that generalizes the Swendsen-Wang method to arbitrary probabilities defined on graph partitions.

Hierarchical learning of curves application to guidewire localization in fluoroscopy

Abstract In this paper we present a method for learning a curve model for detection and segmentat... more Abstract In this paper we present a method for learning a curve model for detection and segmentation by closely integrating a hierarchical curve representation using generative and discriminative models with a hierarchical inference algorithm. We apply this method to the problem of automatic localization of the guidewire in fluoroscopic sequences. In fluoroscopic sequences, the guidewire appears as a hardly visible, non-rigid one-dimensional curve. Our paper has three main contributions.

Hierarchical parsing and semantic navigation of full body CT data

ABSTRACT Whole body CT scanning is a common diagnosis technique for discovering early signs of me... more

Graph partition by Swendsen-Wang cuts

Abstract Vision tasks, such as segmentation, grouping, recognition, can be formulated as graph pa... more Abstract Vision tasks, such as segmentation, grouping, recognition, can be formulated as graph partition problems. The recent literature witnessed two popular graph cut algorithms: the Ncut using spectral graph analysis and the minimum-cut using the maximum flow algorithm. We present a third major approach by generalizing the Swendsen-Wang method-a well celebrated algorithm in statistical mechanics.

Generalizing swendsen-wang for image analysis

Markov chain Monte Carlo (MCMC) methods have been used in many fields (physics, chemistry, biolog... more Markov chain Monte Carlo (MCMC) methods have been used in many fields (physics, chemistry, biology, and computer science) for simulation, inference, and optimization. In many applications, Markov chains are simulated for sampling from target probabilities π (X) defined on graphs G. The graph vertices represent elements of the system, the edges represent spatial relationships, while X is a vector of variables on the vertices which often take discrete values called labels or colors.

Multigrid and multi-level Swendsen-Wang cuts for hierarchic graph partition

Abstract Many vision tasks can be formulated as partitioning an adjacency graph through optimizin... more Abstract Many vision tasks can be formulated as partitioning an adjacency graph through optimizing a Bayesian posterior probability p defined on the partition-space. In this paper two approaches are proposed to generalize the Swendsen-Wang cut algorithm [A. Barbu and SC Zhu 2003] for sampling p. The first method is called multigrid SW-cut which runs SW-cut within a sequence of local" attentional" windows and thus simulates conditional probabilities of p in the partition space.

Incorporating Middle-level Visual Knowledge in Stereo Matching

Multi-Path Marginal Space Learning for Object Detection

Academic Press Library in Signal Processing, 2014

Fast automatic heart chamber segmentation from 3D CT data using marginal space learning and steerable features

Abstract Multi-chamber heart segmentation is a prerequisite for global quantification of the card... more Abstract Multi-chamber heart segmentation is a prerequisite for global quantification of the cardiac function. The complexity of cardiac anatomy, poor contrast, noise or motion artifacts makes this segmentation problem a challenging task. In this paper, we present an efficient, robust, and fully automatic segmentation method for 3D cardiac computed tomography (CT) volumes. Our approach is based on recent advances in learning discriminative object models and we exploit a large database of annotated CT volumes.

RENOIR-A Benchmark Dataset for Real Noise Reduction Evaluation

Hierarchical Object Parsing from Structured Noisy Point Clouds

Eprint Arxiv 1108 3605, Aug 1, 2011

Object parsing and segmentation from point clouds are challenging tasks because the relevant data... more Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as Active Shape and Active Appearance models lack the necessary flexibility for this task, while recent approaches such as the Recursive Compositional Models make model simplifications in order to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer, which is a deformation of a hidden PCA shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state of the art parsing errors on two standard datasets without using any intensity information.

Hierarchical Object Parsing from Structured Noisy Point Clouds

IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013

Object parsing and segmentation from point clouds are challenging tasks because the relevant data... more Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as active shape and active appearance models (AAMs) lack the necessary flexibility for this task, while recent approaches such as the recursive compositional models make model simplifications to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer which is a deformation of a hidden principal component analysis (PCA) shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state-of-the-art parsing errors on two standard datasets without using any intensity information.

Learning mixtures of Bernoulli templates by two-round EM with performance guarantee

Electronic Journal of Statistics, 2014

showed that a two-round variant of the EM algorithm can learn mixture of Gaussian distributions w... more showed that a two-round variant of the EM algorithm can learn mixture of Gaussian distributions with near optimal precision with high probability if the Gaussian distributions are well separated and if the dimension is sufficiently high. In this paper, we generalize their theory to learning mixture of high-dimensional Bernoulli templates. Each template is a binary vector, and a template generates examples by randomly switching its binary components independently with a certain probability. In computer vision applications, a binary vector is a feature map of an image, where each binary component indicates whether a local feature or structure is present or absent within a certain cell of the image domain. A Bernoulli template can be considered as a statistical model for images of objects (or parts of objects) from the same category. We show that the two-round EM algorithm can learn mixture of Bernoulli templates with near optimal precision with high probability, if the Bernoulli templates are sufficiently different and if the number of features is sufficiently high. We illustrate the theoretical results by synthetic and real examples.

Rates for Inductive Learning of Compositional Models

Training an active random field for real-time image denoising

Abstract Many computer vision problems can be formulated in a Bayesian framework based on Markov ... more Abstract Many computer vision problems can be formulated in a Bayesian framework based on Markov random fields (MRF) or conditional random fields (CRF). Generally, the MRF/CRF model is learned independently of the inference algorithm that is used to obtain the final result. In this paper, we observe considerable gains in speed and accuracy by training the MRF/CRF model together with a fast and suboptimal inference algorithm.

SPADES and mixture models

Abstract This paper studies sparse density estimation via ℓ 1 penalization (SPADES). We focus on ... more Abstract This paper studies sparse density estimation via ℓ 1 penalization (SPADES). We focus on estimation in high-dimensional mixture models and nonparametric adaptive density estimation. We show, respectively, that SPADES can recover, with high probability, the unknown components of a mixture of probability densities and that it yields minimax adaptive density estimates. These results are based on a general sparsity oracle inequality that the SPADES estimates satisfy.

Automatic detection and segmentation of axillary lymph nodes

Lymph node detection and measurement is a difficult and important part of cancer treatment. In th... more Lymph node detection and measurement is a difficult and important part of cancer treatment. In this paper we present a robust and effective learning-based method for the automatic detection of solid lymph nodes from Computed Tomography data. The contributions of the paper are the following. First, it presents a learning based approach to lymph node detection based on Marginal Space Learning. Second, it presents an efficient MRF-based segmentation method for solid lymph nodes.

Accurate polyp segmentation for 3D CT colongraphy using multi-staged probabilistic binary learning and compositional model

2008 IEEE Conference on Computer Vision and Pattern Recognition, 2008

Accurate and automatic colonic polyp segmentation and measurement in Computed Tomography (CT) has... more Accurate and automatic colonic polyp segmentation and measurement in Computed Tomography (CT) has significant importance for 3D polyp detection, classification, and more generally computer aided diagnosis of colon cancers. In this paper, we propose a three-staged probabilistic binary classification approach for automatically segmenting polyp voxels from their surrounding tissues in CT. Our system integrates low-, and mid-level information for discriminative learning under local polar coordinates which align on the 3D colon surface around detected polyp. More importantly, our supervised learning system has flexible modeling capacity, which offers a principled means of encoding semantic, clinical expert annotations of colonic polyp tissue identification and segmentation. The learning generality to unseen data is bounded by boosting and stacked generality . Extensive experimental results on polyp segmentation performance evaluation and robustness testing with disturbances (using both training data and unseen data) are provided to validate our presented approach. The reliability of polyp segmentation and measurement has been largely increased to 98.2% (ie. errors ≤ 3mm), compared with other state of art work of about 75% ∼ 80%.

Dimension reduction and variable selection in case control studies via regularized likelihood optimization

Abstract Dimension reduction and variable selection are performed routinely in case-control studi... more Abstract Dimension reduction and variable selection are performed routinely in case-control studies, but the literature on the theoretical aspects of the resulting estimates is scarce. We bring our contribution to this literature by studying estimators obtained via ℓ 1 penalized likelihood optimization. We show that the optimizers of the ℓ 1 penalized retrospective likelihood coincide with the optimizers of the ℓ 1 penalized prospective likelihood.

Simultaneous Detection and Registration for Ileo-Cecal Valve Detection in 3D CT Colonography

Lecture Notes in Computer Science, 2008

Object detection and recognition has achieved a significant progress in recent years. However rob... more Object detection and recognition has achieved a significant progress in recent years. However robust 3D object detection and segmentation in noisy 3D data volumes remains a challenging problem. Localizing an object generally requires its spatial configuration (i.e., pose, size) being aligned with the trained object model, while estimation of an object's spatial configuration is only valid at locations where the object appears. Detecting object while exhaustively searching its spatial parameters, is computationally prohibitive due to the high dimensionality of 3D search space. In this paper, we circumvent this computational complexity by proposing a novel framework capable of incrementally learning the object parameters (IPL) of location, pose and scale. This method is based on a sequence of binary encodings of the projected true positives from the original 3D object annotations (i.e., the projections of the global optima from the global space into the sections of subspaces). The training samples in each projected subspace are labeled as positive or negative, according their spatial registration distances towards annotations as ground-truth. Each encoding process can be considered as a general binary classification problem and is implemented using probabilistic boosting tree algorithm. We validate our approach with extensive experiments and performance evaluations for Ileo-Cecal Valve (ICV) detection in both clean and tagged 3D CT colonography scans. Our final ICV detection system also includes an optional prior learning procedure for IPL which further speeds up the detection.

Generalizing Swendsen-Wang to sampling arbitrary posterior probabilities

Abstract Many vision tasks can be formulated as graph partition problems that minimize energy fun... more Abstract Many vision tasks can be formulated as graph partition problems that minimize energy functions. For such problems, the Gibbs sampler provides a general solution but is very slow, while other methods, such as Ncut and graph cuts are computationally effective but only work for specific energy forms and are not generally applicable. In this paper, we present a new inference algorithm that generalizes the Swendsen-Wang method to arbitrary probabilities defined on graph partitions.

Hierarchical learning of curves application to guidewire localization in fluoroscopy

Abstract In this paper we present a method for learning a curve model for detection and segmentat... more Abstract In this paper we present a method for learning a curve model for detection and segmentation by closely integrating a hierarchical curve representation using generative and discriminative models with a hierarchical inference algorithm. We apply this method to the problem of automatic localization of the guidewire in fluoroscopic sequences. In fluoroscopic sequences, the guidewire appears as a hardly visible, non-rigid one-dimensional curve. Our paper has three main contributions.

Hierarchical parsing and semantic navigation of full body CT data

ABSTRACT Whole body CT scanning is a common diagnosis technique for discovering early signs of me... more

Graph partition by Swendsen-Wang cuts

Abstract Vision tasks, such as segmentation, grouping, recognition, can be formulated as graph pa... more Abstract Vision tasks, such as segmentation, grouping, recognition, can be formulated as graph partition problems. The recent literature witnessed two popular graph cut algorithms: the Ncut using spectral graph analysis and the minimum-cut using the maximum flow algorithm. We present a third major approach by generalizing the Swendsen-Wang method-a well celebrated algorithm in statistical mechanics.

Generalizing swendsen-wang for image analysis

Markov chain Monte Carlo (MCMC) methods have been used in many fields (physics, chemistry, biolog... more Markov chain Monte Carlo (MCMC) methods have been used in many fields (physics, chemistry, biology, and computer science) for simulation, inference, and optimization. In many applications, Markov chains are simulated for sampling from target probabilities π (X) defined on graphs G. The graph vertices represent elements of the system, the edges represent spatial relationships, while X is a vector of variables on the vertices which often take discrete values called labels or colors.

Multigrid and multi-level Swendsen-Wang cuts for hierarchic graph partition

Abstract Many vision tasks can be formulated as partitioning an adjacency graph through optimizin... more Abstract Many vision tasks can be formulated as partitioning an adjacency graph through optimizing a Bayesian posterior probability p defined on the partition-space. In this paper two approaches are proposed to generalize the Swendsen-Wang cut algorithm [A. Barbu and SC Zhu 2003] for sampling p. The first method is called multigrid SW-cut which runs SW-cut within a sequence of local" attentional" windows and thus simulates conditional probabilities of p in the partition space.

Incorporating Middle-level Visual Knowledge in Stereo Matching

Multi-Path Marginal Space Learning for Object Detection

Academic Press Library in Signal Processing, 2014

Fast automatic heart chamber segmentation from 3D CT data using marginal space learning and steerable features

Abstract Multi-chamber heart segmentation is a prerequisite for global quantification of the card... more Abstract Multi-chamber heart segmentation is a prerequisite for global quantification of the cardiac function. The complexity of cardiac anatomy, poor contrast, noise or motion artifacts makes this segmentation problem a challenging task. In this paper, we present an efficient, robust, and fully automatic segmentation method for 3D cardiac computed tomography (CT) volumes. Our approach is based on recent advances in learning discriminative object models and we exploit a large database of annotated CT volumes.

On the range of non-vanishing p-torsion cohomology for GLn (Fp)

The range of non-vanishing of H

The Ring Generated by the Elements of Degree 2 in H*(Un(p),)

Journal of Algebra, 2001

We compute all the relations in cohomology satisfied by the elements of degree two of H*(Un(p), )... more

On the cohomology of GLn (Fp) with Fp coefficients

ABSTRACT In this thesis we will develop some tools that can be used to get a better understanding... more ABSTRACT In this thesis we will develop some tools that can be used to get a better understanding of the cohomology of GLn (Fp) with Fp coefficients. First we study the cohomology of Un (Fp), one of the p-Sylow subgroups of GLn (Fp). We compute all the relations satisfied by those elements of H∗(Un) of degree two which come from the Z cohomology.

On a conjecture of Ash

In this paper we prove a particular case of a Conjecture of Ash that states the existence of Galo... more

Learning-based Component for Suppression of False Positives Located on the Ileo Cecal Valve: Evaluation of Performance on 802 CTC Volumes

PURPOSE Evaluate the performance of a module for suppression of false positive CAD marks located ... more PURPOSE Evaluate the performance of a module for suppression of false positive CAD marks located on the ileo cecal valve (ICV) in a prototype polyp detection system when applied to clean or tagged CTC datasets METHOD AND MATERIALS The ICV detection component uses 5 steps: a) detecting the ICV orifice leveraging its distinctive local curvature profile, b) a rough estimation of its orientation by aligning it with the local gradients at a given location. Estimation of the c) position, d) size of the ICV and e) refining the orientation is calculated by performing marginal space learning. In each of the 5 steps, all potential orifice candidates are evaluated using different classifiers. The top 100 candidates with maximal orifice probabilities are selected for further parameter estimation and searching. The number of selected candidates is set to maintain a good trade-off between detection accuracy and speed. The position of 116 manually marked ICV orifices were used to generate a traini...

Learning-based Component for Suppression Rectal Tube False Positives: Evaluation of Performance on 780 CTC Cases

Learning-based Component for Suppression of False Positives Located on the Ileo Cecal Valve: Evaluation of Performance on 802 CTC Volumes

ABSTRACT PURPOSE Evaluate the performance of a module for suppression of false positive CAD marks... more ABSTRACT PURPOSE Evaluate the performance of a module for suppression of false positive CAD marks located on the ileo cecal valve (ICV) in a prototype polyp detection system when applied to clean or tagged CTC datasets METHOD AND MATERIALS The ICV detection component uses 5 steps: a) detecting the ICV orifice leveraging its distinctive local curvature profile, b) a rough estimation of its orientation by aligning it with the local gradients at a given location. Estimation of the c) position, d) size of the ICV and e) refining the orientation is calculated by performing marginal space learning. In each of the 5 steps, all potential orifice candidates are evaluated using different classifiers. The top 100 candidates with maximal orifice probabilities are selected for further parameter estimation and searching. The number of selected candidates is set to maintain a good trade-off between detection accuracy and speed. The position of 116 manually marked ICV orifices were used to generate a training set of 8,360 positive and 3,700,000 negative samples being used to train a probabilistic boosting tree. On the training set a sensitivity of 97.4% (113/116) was achieved. The processing time varied from 4 to 10 seconds on a P4 3.2GHz computer RESULTS The ICV detection was tested on 802 unseen datasets: 407 clean volumes from 10 different sites; 395 tagged volumes, including iodine and barium preparations, from 2 sites. The ICV detection was implemented as post filter. In clean cases, ICV detection reduced the number of false positives (fp) from 3.92 fp/pat (2.04 fp/vol.) to 3.72 fp/pat (1.92 fp/vol.) without impacting the overall sensitivity of the CAD system. In tagged cases ICV detection reduced the number of false marks from 6.2 fp/pat (3.15 fp/vol.) to 5.78 fp/pat (2.94 fp/vol.). 1 polyp out of 121 polyps with a size range from 6 up to 25 mm was wrongly labeled as ICV, resulting in a sensitivity drop of 0.8%. CONCLUSION ICV detection effectively reduces detection false positives, both in clean and tagged cases, from the list of marks presented to the reader by ColonCAD without impacting the system sensitivity CLINICAL RELEVANCE/APPLICATION Ileo cecal valve detection and false positive reduction improve the performance and acceptance of CAD systems for CTC

Methods and apparatus for virtual coronary mapping

A virtual map of vessels of interest in medical procedures, such as coronary angioplasty is creat... more A virtual map of vessels of interest in medical procedures, such as coronary angioplasty is created so that doses of contrasting agent given to a patient may be reduced. A position of a coronary guidewire is determined and locations of vessel boundaries are found. When the contrast agent has dissipated, virtual maps of the vessels are created as new images. The locations of the determined vessel boundaries are imported to a mapping system and an image obtained without using a contrast agent is modified based on the imported ...

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

The present invention is directed to a method for populating a database with a set of images of a... more 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 ...

Method and system for generating a four-chamber heart model

A method and system for building a statistical four-chamber heart model from 3D volumes is disclo... more A method and system for building a statistical four-chamber heart model from 3D volumes is disclosed. In order to generate the four-chamber heart model, each chamber is modeled using an open mesh, with holes at the valves. Based on the image data in one or more 3D volumes, meshes are generated and edited for the left ventricle (LV), left atrium (LA), right ventricle (RV), and right atrium (RA). Resampling to enforce point correspondence is performed during mesh editing. Important anatomic landmarks in the heart are explicitly ...

Method for detecting polyps in a three dimensional image volume

A method for detecting target objects in a three dimensional (3D) image volume of an anatomical s... more A method for detecting target objects in a three dimensional (3D) image volume of an anatomical structure is disclosed. A set of candidate locations in the image volume are obtained. For each candidate location, sub-volumes of at least two different scales are cropped out. Each sub-volume comprises a plurality of voxels. For each of the sub-volumes, each sub-volume is rotated in at least two different orientations. A shape classifier is applied to each sub-volume. If the voxels in the sub-volume pass the shape classifier, a gradient ...

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

A system and method for using learned discriminative models to segment a border of an anatomical ... more 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 XY planes are obtained. Seed regions are selected in the 2D slices. Morphological region growing is performed on the selected seed regions. An initial ...

System and method for segmenting chambers of a heart in a three dimensional image

A system and method for segmenting chambers of a heart in three dimensional images is disclosed. ... more A system and method for segmenting chambers of a heart in three dimensional images is disclosed. A set of three dimensional images of a heart is received. The shape of the heart in the three dimensional images is localized. Boundaries of the chambers of the heart in the localized shape are identified using steerable features.

System and method for detecting and tracking a guidewire in a fluoroscopic image sequence

A system and method for populating a database with a set of image sequences of an object is discl... more A system and method for populating a database with a set of image sequences of an object is disclosed. The database is used to detect localization of a guidewire in the object. A set of images of anatomical structures is received in which each image is annotated to show a guidewire, catheter, wire tip and stent. For each given image a Probabilistic Boosting Tree (PBT) is used to detect short line segments of constant length in the image. Two segment curves are constructed from the short line segments. A discriminative joint shape and ...

System and method for detecting an object in a high dimensional space

A system and method for detecting an object in a high dimensional image space is disclosed. A thr... more A system and method for detecting an object in a high dimensional image space is disclosed. A three dimensional image of an object is received. A first classifier is trained in the marginal space of the object center location which generates a predetermined number of candidate object center locations. A second classifier is trained to identify potential object center locations and orientations from the predetermined number of candidate object center locations and maintaining a subset of the candidate object center locations. A third ...

System and method for coronary digital subtraction angiography

A method and system for extracting motion-based layers from fluoroscopic image sequences are disc... more A method and system for extracting motion-based layers from fluoroscopic image sequences are disclosed. Portions of multiple objects, such as anatomical structures, are detected in the fluoroscopic images. Motion of the objects is estimated between the images is the sequence of fluoroscopic images. The images in the fluoroscopic image sequence are then divided into layers based on the estimated motion. In a particular implementation, the coronary vessel tree and the diaphragm can be extracted in separate motion layers from coronary ...

Method and system for catheter detection and tracking in a fluoroscopic image sequence

by Norbert Strobel and Adrian Barbu

A method and system for detecting and tracking an ablation catheter tip in a fluoroscopic image s... more A method and system for detecting and tracking an ablation catheter tip in a fluoroscopic image sequence is disclosed. Catheter tip candidates are detected in each frame of the fluoroscopic image sequence using marginal space learning. The detected catheter tip candidates are then tracked over all the frames of the fluoroscopic image sequence in order to determine an ablation catheter tip location in each frame.

Method and system for object detection using probabilistic boosting cascade tree

A method and system for object detection using a probabilistic boosting cascade tree (PBCT) is di... more A method and system for object detection using a probabilistic boosting cascade tree (PBCT) is disclosed. A PBCT is a machine learning based classifier having a structure that is driven by training data and determined during the training process without user input. In a PBCT training method, for each node in the PBCT, a classifier is trained for the node based on training data received at the node. The performance of the classifier trained for the node is then evaluated based on the training data. Based on the performance of the classifier, ...

Method and system for polyp segmentation for 3D computed tomography colonography

by Luca Bogoni and Adrian Barbu

A method and system for polyp segmentation in computed tomography colonogrphy (CTC) volumes is di... more A method and system for polyp segmentation in computed tomography colonogrphy (CTC) volumes is disclosed. The polyp segmentation method utilizes a three-staged probabilistic binary classification approach for automatically segmenting polyp voxels from surrounding tissue in CTC volumes. Based on an input initial polyp position, a polyp tip is detected in a CTC volume using a trained 3D point detector. A local polar coordinate system is then fit to the colon surface in the CTC volume with the origin at the detected polyp tip. Polyp interior ...

User interface for polyp annotation, segmentation, and measurement in 3D computed tomography colonography

A method and system for providing a user interface for polyp annotation, segmentation, and measur... more A method and system for providing a user interface for polyp annotation, segmentation, and measurement in computer tomography colonography (CTC) volumes is disclosed. The interface receives an initial polyp position in a CTC volume, and automatically segments the polyp based on the initial polyp position. In order to segment the polyp, a polyp tip is detected in the CTC volume using a trained 3D point detector. A local polar coordinate system is then fit to the colon surface in the CTC volume with the origin at the detected ...

Method and system for vessel segmentation in fluoroscopic images

A method and system for vessel segmentation in fluoroscopic images is disclosed. Hierarchical lea... more A method and system for vessel segmentation in fluoroscopic images is disclosed. Hierarchical learning-based detection is used to perform the vessel segmentation. A boundary classifier is trained and used to detect boundary pixels of a vessel in a fluoroscopic image. A cross-segment classifier is trained and used to detect cross-segments connecting the boundary pixels. A quadrilateral classifier is trained and used to detect quadrilaterals connecting the cross segments. Dynamic programming is then used to ...

System and method for coronary digital subtraction angiography

A method and system for extracting coronary vessels fluoroscopic image sequences using coronary d... more A method and system for extracting coronary vessels fluoroscopic image sequences using coronary digital subtraction angiography (DSA) are disclosed. A set of mask images of a coronary region is received, and a sequence of contrast images for the coronary region is received. For each contrast image, vessel regions are detected in the contrast image using learning-based vessel segment detection and a background region of the contrast image is determined based on the detected vessel regions. Background motion is estimated ...

Method and system for evaluating image segmentation based on visibility

A method and system for evaluating image segmentation is disclosed. In order to quantitatively ev... more A method and system for evaluating image segmentation is disclosed. In order to quantitatively evaluate an image segmentation technique, synthetic image data is generated and the synthetic image data is segmented to extract an object using the segmentation technique. This segmentation results in a foreground containing the extracted object and a background. The visibility of the extracted object is quantitatively measured based on the intensity distributions of the segmented foreground and background. The visibility is ...

Method and system for detection and registration of 3D objects using incremental parameter learning

A method and system for detecting 3D objects in images is disclosed. In particular, a method and ... more A method and system for detecting 3D objects in images is disclosed. In particular, a method and system for Ileo-Cecal Valve detection in 3D computed tomography (CT) images using incremental parameter learning and ICV specific prior learning is disclosed. First, second, and third classifiers are sequentially trained to detect candidates for position, scale, and orientation parameters of a box that bounds an object in 3D image. In the training of each sequential classifier, new training samples are generated by scanning the object's ...

Hierarchical Object Parsing from Noisy Point Clouds

Arxiv preprint arXiv:1108.3605, 2011

Abstract: Object parsing and segmentation from point clouds are challenging tasks because the rel... more

Face Detection Using a 3D Model on Face Keypoints

This paper presents a part-based face detection approach where the spatial relationship between t... more This paper presents a part-based face detection approach where the spatial relationship between the face parts is represented by a 3D model with six parameters. The computational complexity of the search in the six dimensional pose space is addressed by proposing meaningful 3D pose candidates by image-based regression, and scoring them using a parameter sensitive classifier that uses features at locations specified by the 3D pose of the candidate. A compatible subset of candidates is then obtained by non-maximal suppression. Experiments on two standard face detection datasets show that the proposed 3D model based approach obtains results comprable to state of the art.

Fast Simultaneous Feature Selection and Learning

Many learning problems need the selection of a given number of features from a larger pool and a ... more Many learning problems need the selection of a given number of features from a larger pool and a classifier trained on the selected features. Penalized methods based on sparsity inducing priors can be slow on large datasets, with millions of observations and features. Moreover, the sparsity priors might introduce a bias in the estimated parameters and a second step is often used to learn an unbiased model. In this paper we propose a novel efficient algorithm that simultaneously selects a desired number of variables and learns a model on them, by optimizing a likelihood with a sparsity constraint. The iterative algorithm alternates parameter updates with tightening the sparsity constraint by gradually removing variables based on a criterion and a schedule. We present a generic approach for optimizing any differentiable loss function and an application to logistic regression with parametric and non-parametric formulations and consistency guarantees. Experiments on real and simulated data show that the proposed method outperforms state of the art methods based on sparsity inducing penalties for both variable selection and prediction while being computationally faster.

Hierarchical Object Parsing from Noisy Point Clouds

Object parsing and segmentation from point clouds are challenging tasks because the relevant data... more Object parsing and segmentation from point clouds are challenging tasks because the relevant data is available only as thin structures along object boundaries or other features, and is corrupted by large amounts of noise. To handle this kind of data, flexible shape models are desired that can accurately follow the object boundaries. Popular models such as Active Shape and Active Appearance models lack the necessary flexibility for this task, while recent approaches such as the Recursive Compositional Models make model simplifications in order to obtain computational guarantees. This paper investigates a hierarchical Bayesian model of shape and appearance in a generative setting. The input data is explained by an object parsing layer, which is a deformation of a hidden PCA shape model with Gaussian prior. The paper also introduces a novel efficient inference algorithm that uses informed data-driven proposals to initialize local searches for the hidden variables. Applied to the problem of object parsing from structured point clouds such as edge detection images, the proposed approach obtains state of the art parsing errors on two standard datasets without using any intensity information.

Face Detection Using a 3D Model on Face Keypoints

This paper presents a part-based face detection approach where the spatial relationship between t... more This paper presents a part-based face detection approach where the spatial relationship between the face parts is represented by a 3D model with six parameters. The computational complexity of the search in the six dimensional pose space is addressed by proposing meaningful 3D pose candidates by image-based regression, and scoring them using a parameter sensitive classifier that uses features at locations specified by the 3D pose of the candidate. A compatible subset of candidates is then obtained by non-maximal suppression. Experiments on two standard face detection datasets show that the proposed 3D model based approach obtains results comprable to state of the art.

Fast Simultaneous Feature Selection and Learning

Many computer vision and medical imaging problems are faced with learning from large-scale datase... more Many computer vision and medical imaging problems are faced with learning from large-scale datasets, with millions of observations and features. In this paper we propose a novel efficient learning scheme that tightens a sparsity constraint by gradually removing variables based on a criterion and a schedule. The attractive fact that the problem size keeps dropping throughout the iterations makes it particularly suitable for big data learning. Our approach applies generically to the optimization of any differentiable loss function, and finds applications in regression, classification and ranking. The resultant algorithms build variable screening into estimation and are extremely simple to implement. We provide theoretical guarantees of convergence and selection consistency. In addition, one dimensional piecewise linear response functions are used to account for nonlinearity and a second order prior is imposed on these functions to avoid overfitting. Experiments on real and synthetic data show that the proposed method compares very well with other state of the art methods in regression, classification and ranking while being computationally very efficient and scalable.

Feature Selection with Annealing for Big Data Learning

Many computer vision and medical imaging problems are faced with learning from large-scale datase... more Many computer vision and medical imaging problems are faced with learning from large-scale datasets, with millions of observations and features. In this paper we propose a novel efficient learning scheme that tightens a sparsity constraint by gradually removing variables based on a criterion and a schedule. The attractive fact that the problem size keeps dropping throughout the iterations makes it particularly suitable for big data learning. Our approach applies generically to the optimization of any differentiable loss function, and finds applications in regression, classification and ranking. The resultant algorithms build variable screening into estimation and are extremely simple to implement. We provide theoretical guarantees of convergence and selection consistency. In addition, one dimensional piecewise linear response functions are used to account for nonlinearity and a second order prior is imposed on these functions to avoid overfitting. Experiments on real and synthetic data show that the proposed method compares very well with other state of the art methods in regression, classification and ranking while being computationally very efficient and scalable.