A5012751147- Medical Image Segmentation Techniques
- Radiomics and Machine Learning in Medical Imaging
- Cardiac Valve Diseases and Treatments
- AI in cancer detection
- COVID-19 diagnosis using AI
- Advanced Neural Network Applications
- Medical Imaging and Analysis
- Cardiovascular Function and Risk Factors
- Lung Cancer Diagnosis and Treatment
- Advanced Vision and Imaging
- Image Retrieval and Classification Techniques
- Coronary Interventions and Diagnostics
- Medical Imaging Techniques and Applications
- Cardiac Imaging and Diagnostics
- Advanced Image and Video Retrieval Techniques
- Video Surveillance and Tracking Methods
- Advanced X-ray and CT Imaging
- Advanced MRI Techniques and Applications
- Cardiac electrophysiology and arrhythmias
- Image and Signal Denoising Methods
- Domain Adaptation and Few-Shot Learning
- Infective Endocarditis Diagnosis and Management
- 3D Shape Modeling and Analysis
- Human Pose and Action Recognition
- Dental Radiography and Imaging
Siemens Healthcare (United States)
2016-2025
Siemens (United States)
2015-2024
Siemens Healthcare (Germany)
2010-2024
Loyola University Medical Center
2024
Radboud University Nijmegen
2024
Radboud University Medical Center
2024
Siemens (Germany)
2004-2020
Princeton University
2007-2018
Istituto Superiore di Sanità
2016
Institute of Electrical and Electronics Engineers
2006
A general non-parametric technique is proposed for the analysis of a complex multimodal feature space and to delineate arbitrarily shaped clusters in it. The basic computational module an old pattern recognition procedure: mean shift. For discrete data, we prove convergence recursive shift procedure nearest stationary point underlying density function and, thus, its utility detecting modes density. relation Nadaraya-Watson estimator from kernel regression robust M-estimators; location also...
A new approach toward target representation and localization, the central component in visual tracking of nonrigid objects, is proposed. The feature histogram-based representations are regularized by spatial masking with an isotropic kernel. induces spatially-smooth similarity functions suitable for gradient-based optimization, hence, localization problem can be formulated using basin attraction local maxima. We employ a metric derived from Bhattacharyya coefficient as measure, use mean...
A new method for real time tracking of non-rigid objects seen from a moving camera is proposed. The central computational module based on the mean shift iterations and finds most probable target position in current frame. dissimilarity between model (its color distribution) candidates expressed by metric derived Bhattacharyya coefficient. theoretical analysis approach shows that it relates to Bayesian framework while providing practical, fast efficient solution. capability tracker handle...
A nonparametric estimator of density gradient, the mean shift, is employed in joint, spatial-range (value) domain gray level and color images for discontinuity preserving filtering image segmentation. Properties shift are reviewed its convergence on lattices proven. The proposed method associates with each pixel closest local mode distribution joint domain. Segmentation into a piecewise constant structure requires only one more step, fusion regions associated nearby modes. technique has two...
A general technique for the recovery of significant image features is presented. The based on mean shift algorithm, a simple nonparametric procedure estimating density gradients. Drawbacks current methods (including robust clustering) are avoided. Feature space any nature can be processed, and as an example, color segmentation discussed. completely autonomous, only its class chosen by user. Thus, same program produce high quality edge image, or provide, extracting all colors, preprocessor...
We propose an automatic four-chamber heart segmentation system for the quantitative functional analysis of from cardiac computed tomography (CT) volumes. Two topics are discussed: modeling and model fitting to unseen volume. Heart is a nontrivial task since complex nonrigid organ. The must be anatomically accurate, allow manual editing, provide sufficient information guide detection segmentation. Unlike previous work, we explicitly represent important landmarks (such as valves ventricular...
In this paper, we present the logarithmic total variation (LTV) model for face recognition under varying illumination, including natural lighting conditions, where rarely know strength, direction, or number of light sources. The proposed LTV has ability to factorize a single image and obtain illumination invariant facial structure, which is then used recognition. Our inspired by SQI but better edge-preserving simpler parameter selection. merit that neither does it require any assumption nor...
We present two solutions for the scale selection problem in computer vision. The first one is completely nonparametric and based on adaptive estimation of normalized density gradient. Employing sample point estimator, we define Variable Bandwidth Mean Shift, prove its convergence, show superiority over fixed bandwidth procedure. second technique has a semiparametric nature imposes local structure data to extract reliable information. underlying taken as which maximizes magnitude mean shift...
Fractional flow reserve (FFR) is a functional index quantifying the severity of coronary artery lesions and clinically obtained using an invasive, catheter-based measurement. Recently, physics-based models have shown great promise in being able to noninvasively estimate FFR from patient-specific anatomical information, e.g., computed tomography scans heart arteries. However, these high computational demand, limiting their clinical adoption. In this paper, we present machine-learning-based...
Robust and fast detection of anatomical structures is a prerequisite for both diagnostic interventional medical image analysis. Current solutions anatomy are typically based on machine learning techniques that exploit large annotated databases in order to learn the appearance captured anatomy. These subject several limitations, including use suboptimal feature engineering most importantly computationally search-schemes detection. To address these issues, we propose method follows new...
3-D image registration, which involves aligning two or more images, is a critical step in variety of medical applications from diagnosis to therapy. Image registration commonly performed by optimizing an matching metric as cost function. However this task challenging due the non-convex nature over plausible parameter space and insufficient approches for robust optimization. As result, current approaches are often customized specific problem sensitive quality artifacts. In paper, we propose...
Purpose To present a method that automatically segments and quantifies abnormal CT patterns commonly in COVID-19, namely ground-glass opacities consolidations. Materials Methods In this retrospective study, the proposed takes as input noncontrast chest lesions, lungs, lobes three dimensions, based on dataset of 9749 volumes. The outputs two combined measures severity lung lobe involvement, quantifying both extent COVID-19 abnormalities presence high opacities, deep learning reinforcement...
The analysis of a feature space that exhibits multiscale patterns often requires kernel estimation techniques with locally adaptive bandwidths, such as the variable-bandwidth mean shift. Proper selection bandwidth is, however, critical step for superior and partitioning. This paper presents shift-based approach local in multimodal, multivariate case. method is based on fundamental property normal distributions regarding bias normalized density gradient. demonstrates that, within large sample...
A new paradigm for the efficient color-based tracking of objects seen from a moving camera is presented. The proposed technique employs mean shift analysis to derive target candidate that most similar given model, while prediction next location computed with Kalman filter. dissimilarity between model and candidates expressed by metric based on Bhattacharyya coefficient. implementation method achieves real-time performance, being appropriate large variety different color patterns. resulting...
We propose a novel method for the automatic detection and measurement of fetal anatomical structures in ultrasound images. This problem offers myriad challenges, including: difficulty modeling appearance variations visual object interest, robustness to speckle noise signal dropout, large search space procedure. Previous solutions typically rely on explicit encoding prior knowledge formulation as perceptual grouping task solved through clustering or variational approaches. These methods are...
As decisions in cardiology increasingly rely on noninvasive methods, fast and precise image processing tools have become a crucial component of the analysis workflow. To best our knowledge, we propose first automatic system for patient-specific modeling quantification left heart valves, which operates cardiac computed tomography (CT) transesophageal echocardiogram (TEE) data. Robust algorithms, based recent advances discriminative learning, are used to estimate parameters from sequences...
Visual features are commonly modeled with probability density functions in computer vision problems, but current methods such as a mixture of Gaussians and kernel estimation suffer from either the lack flexibility, by fixing or limiting number Gaussian components mixture, large memory requirement, maintaining non-parametric representation density. These problems aggravated real-time applications since required to be updated new data becomes available. We present novel approximation technique...