A5041274548- Cell Image Analysis Techniques
- AI in cancer detection
- Medical Image Segmentation Techniques
- Image and Signal Denoising Methods
- Anomaly Detection Techniques and Applications
- Digital Imaging for Blood Diseases
- Advanced Neuroimaging Techniques and Applications
- COVID-19 Clinical Research Studies
- Retinal Imaging and Analysis
- Advanced MRI Techniques and Applications
- Advanced Numerical Analysis Techniques
- Gene expression and cancer classification
- Morphological variations and asymmetry
- Medical Imaging Techniques and Applications
- Advanced Image Processing Techniques
- Platelet Disorders and Treatments
- Neural Networks and Applications
- Advanced Fluorescence Microscopy Techniques
- Advanced Vision and Imaging
- Facial Nerve Paralysis Treatment and Research
- Radiomics and Machine Learning in Medical Imaging
- Adversarial Robustness in Machine Learning
- Heparin-Induced Thrombocytopenia and Thrombosis
- Topological and Geometric Data Analysis
- Human Pose and Action Recognition
University of Virginia
2016-2025
University of California, Davis
2024
University of North Carolina at Chapel Hill
2024
North South University
2024
National Institutes of Health Clinical Center
2024
Carnegie Mellon University
2008-2021
The University of Tokyo
2021
HRL Laboratories (United States)
2020
Institute of Electrical and Electronics Engineers
2020
Signal Processing (United States)
2020
Abstract Patient motion and image distortion induced by eddy currents cause artifacts in maps of diffusion parameters computed from diffusion‐weighted (DW) images. A novel comprehensive approach to correct for spatial misalignment DW imaging (DWI) volumes acquired with different strengths orientations the sensitizing gradients is presented. This uses a mutual information‐based registration technique transformation model containing that current‐induced rigid body three dimensions. All are...
To characterize anisotropic water diffusion in brain white matter, a theoretical framework is proposed that combines hindered and restricted models of (CHARMED) an experimental methodology embodies features tensor q-space MRI. This model contains extra-axonal compartment, whose properties are characterized by effective tensor, intra-axonal within cylinders. The primarily explains the Gaussian signal attenuation observed at low b values; non-Gaussian does so high b. Both data obtained along...
Nonrigid registration of medical images is important for a number applications such as the creation population averages, atlas-based segmentation, or geometric correction functional magnetic resonance imaging (IMRI) to name few. In recent years, methods have been proposed solve this problem, one class which involves maximizing mutual information (Ml)-based objective function over regular grid splines. This approach has produced good results but its computational complexity proportional...
Transport-based techniques for signal and data analysis have recently received increased interest. Given their ability to provide accurate generative models intensities other distributions, they been used in a variety of applications, including content-based retrieval, cancer detection, image superresolution, statistical machine learning, name few, shown produce state-of-the-art results. Moreover, the geometric characteristics transport-related metrics inspired new kinds algorithms...
Images taken under water usually suffer from the problems of quality degradation, such as low contrast, blurring details, color deviations, non-uniform illumination, etc. As an important problem in image processing and computer vision, restoration enhancement underwater are necessary for numerous practical applications. Over last few decades, have been attracting increasing amount research effort. However, a comprehensive in-depth survey related achievements improvements is still missing,...
Optimal transport distances, otherwise known as Wasserstein have recently drawn ample attention in computer vision and machine learning powerful discrepancy measures for probability distributions. The recent developments on alternative formulations of the optimal allowed faster solutions to problem revamped their practical applications learning. In this paper, we exploit widely used kernel methods provide a family provably positive definite kernels based Sliced distance demonstrate benefits...
Gaussian mixture models (GMM) are powerful parametric tools with many applications in machine learning and computer vision. Expectation maximization (EM) is the most popular algorithm for estimating GMM parameters. However, EM guarantees only convergence to a stationary point of log-likelihood function, which could be arbitrarily worse than optimal solution. Inspired by relationship between negative function Kullback-Leibler (KL) divergence, we propose an alternative formulation parameters...
Extracting high-resolution information from highly degraded facial images is an important problem with several applications in science and technology. Here we describe a single frame super resolution technique that uses transport-based formulation of the problem. The method consists training testing phase. In phase, nonlinear Lagrangian model appearance constructed fully automatically. image enhanced by finding parameters best fit given low data. We test approach on two face datasets, namely...
A characteristic clinical feature of COVID-19 is the frequent incidence microvascular thrombosis. In fact, autopsy reports have shown widespread thrombotic microangiopathy characterized by extensive diffuse microthrombi within peripheral capillaries and arterioles in lungs, hearts, other organs, resulting multiorgan failure. However, underlying process COVID-19-associated thrombosis remains elusive due to lack tools statistically examine platelet aggregation (i.e., initiation microthrombus...
Follicular lesions of the thyroid are traditionally difficult and tedious challenges in diagnostic surgical pathology part due to lack obvious discriminatory cytological microarchitectural features. We describe a computerized method detect classify follicular adenoma thyroid, carcinoma normal based on nuclear chromatin distribution from digital images tissue obtained by routine histological methods. Our is determining whether set nuclei, using automated image segmentation, most similar sets...
Invertible image representation methods (transforms) are routinely employed as low-level processing operations based on which feature extraction and recognition algorithms developed. Most transforms in current use (e.g. Fourier, Wavelet, etc.) linear transforms, and, by themselves, unable to substantially simplify the of classes for classification. Here we describe a nonlinear, invertible, transform combining well known Radon data, 1D Cumulative Distribution Transform proposed earlier. We...
We describe a simple but robust algorithm for estimating the heart rate pulse from video sequences containing human skin in real time. Based on model of light interaction with skin, we define change blood concentration due to arterial pulsation as pixel quotient log space, and successfully use derived signal computing rate. Various experiments different cameras, illumination condition, locations were conducted demonstrate effectiveness robustness proposed algorithm. Examples computed normal...
Significance Much of what is currently known about how cells work has been derived through visual interpretation microscopy images. Computational methods for image analysis have emerged as quantitative alternatives to interpretation. We describe an pipeline cell databases that combines statistical pattern recognition with the mathematics optimal mass transport. The approach fully automated and does not require use ad hoc numerical features. It enables identification discriminant phenotypic...
Abstract We describe a new supervised learning‐based template matching approach for segmenting cell nuclei from microscopy images. The method uses examples selected by user building statistical model that captures the texture and shape variations of nuclear structures given dataset to be segmented. Segmentation subsequent, unlabeled, images is then performed finding instance best matches (in normalized cross correlation sense) local neighborhood in input image. demonstrate application our...
Epithelium-stroma classification is a necessary preprocessing step in histopathological image analysis. Current deep learning based recognition methods for histology data require collection of large volumes labeled order to train new neural network when there are changes the acquisition procedure. However, it extremely expensive pathologists manually label sufficient each pathology study professional manner, which results limitations real-world applications. A very simple but effective...
In this paper we study generative modeling via autoencoders while using the elegant geometric properties of optimal transport (OT) problem and Wasserstein distances. We introduce Sliced-Wasserstein Autoencoders (SWAE), which are models that enable one to shape distribution latent space into any samplable probability without need for training an adversarial network or defining a closed-form distribution. short, regularize autoencoder loss with sliced-Wasserstein distance between encoded...
Free space optical communications utilizing orbital angular momentum beams have recently emerged as a new technique for with potential increased channel capacity. Turbulence due to changes in the index of refraction emanating from temperature, humidity, and air flow patterns, however, add nonlinear effects received thus making demultiplexing task more difficult. Deep learning techniques been previously applied solve problem an image classification task. Here we make use newly developed...
The automatic classification of breast cancer histopathological images has great significance in computer-aided diagnosis. Recently, deep learning via neural networks enabled pattern detection and prediction using large, labeled datasets; whereas, collecting annotating sufficient histological data professional pathologists is time consuming, tedious, extremely expensive. In the proposed paper, a active framework designed implemented for images, with goal maximizing accuracy from very limited...
Autism is traditionally diagnosed behaviorally but has a strong genetic basis. A genetics-first approach could transform understanding and treatment of autism. However, isolating the gene-brain-behavior relationship from confounding sources variability challenge. We demonstrate novel technique, 3D transport-based morphometry (TBM), to extract structural brain changes linked copy number variation (CNV) at 16p11.2 region. identified two distinct endophenotypes. In data Simons Variation in...