A5060672357- Medical Image Segmentation Techniques
- Medical Imaging Techniques and Applications
- Topological and Geometric Data Analysis
- Data Visualization and Analytics
- Neural Networks and Applications
- Advanced Neuroimaging Techniques and Applications
- Functional Brain Connectivity Studies
- Model Reduction and Neural Networks
- Sparse and Compressive Sensing Techniques
- Domain Adaptation and Few-Shot Learning
- Gaussian Processes and Bayesian Inference
- Cell Image Analysis Techniques
- Face and Expression Recognition
- Image Retrieval and Classification Techniques
- Image and Signal Denoising Methods
- Advanced Neural Network Applications
- Advanced Mathematical Modeling in Engineering
- Morphological variations and asymmetry
- Medical Imaging and Analysis
- Radiomics and Machine Learning in Medical Imaging
- COVID-19 diagnosis using AI
- Criminal Law and Evidence
- Point processes and geometric inequalities
- Cerebral Venous Sinus Thrombosis
- Machine Learning and Algorithms
Kitware (United States)
2017-2022
Duke University
2013
University of Utah
2007-2012
An important goal of scientific data analysis is to understand the behavior a system or process based on sample system. In many instances it possible observe both input parameters and outputs, characterize as high-dimensional function. Such sets arise, for instance, in large numerical simulations, energy landscapes optimization problems, image relating biological medical parameters. This paper proposes an approach analyze visualizing such sets. The proposed method combines topological...
The degree of correlation between variables is used in many data analysis applications as a key measure interdependence. most common techniques for exploratory pairwise multivariate datasets, like scatterplot matrices and clustered heatmaps, however, do not scale well to large either computationally or visually. We present new visualization that capable encoding hundreds thousands variables, called the s-CorrPlot. s-CorrPlot encodes spatially points on using geometric structure underlying...
As dataset size and complexity steadily increase, uncertainty is becoming an important data aspect. So, today's visualizations need to incorporate indications of uncertainty. However, characterizing for visualization isn't always straightforward. Entropy, in the information-theoretic sense, can be a measure categorical datasets. The authors discuss mathematical formulation, interpretation, use entropy visualizations. This research aims demonstrate as metric expand vocabulary measures visualization.
Non-linear dimensionality reduction of noisy data is a challenging problem encountered in variety analysis applications. Recent results the literature show that spectral decomposition, as used for example by Laplacian Eigenmaps algorithm, provides powerful tool non-linear and manifold learning. In this paper, we discuss significant shortcoming these approaches, which refer to repeated eigendirections problem. We propose novel approach combines successive 1-dimensional embeddings with...
This article introduces a novel partition-based regression approach that incorporates topological information. Partition-based typically quality-of-fit-driven decomposition of the domain. The emphasis in this work is on topologically meaningful segmentation. Thus, proposed based segmentation induced by discrete approximation Morse–Smale complex. yields with partitions corresponding to regions function single minimum and maximum are often well approximated linear model. models amenable...
In many areas, scientists deal with increasingly high-dimensional data sets. An important aspect for these is to gain a qualitative understanding of the process or system from which gathered. Often, both input variables and an outcome are observed can be characterized as sample scalar function. This work presents R package <b>msr</b> exploratory analysis multivariate functions based on Morse-Smale complex. The complex provides topologically meaningful decomposition domain. implements...
We present a manifold learning approach to dimensionality reduction that explicitly models the as mapping from low high dimensional space. The is represented parametrized surface by set of parameters are defined on input samples. representation also provides natural space, and concatenation these two mappings induces projection operator onto manifold. explicit allows for clearly objective function in terms distance reconstruction error. A formulation kernel regression permits direct...
This paper presents a multiscale approach to efficiently compute approximate optimal transport plans between point sets. It is particularly well-suited for sets that are in high-dimensions, but close being intrinsically low-dimensional. The based on an adaptive decomposition of the yields sequence problems, solved top-to-bottom fashion from coarsest finest scale. We provide numerical evidence this scales approximately linearly, time and memory, number nodes, instead quadratically or worse...
Ultrasound is an effective tool for rapid noninvasive assessment of cardiac structure and function. Determining the cardiorespiratory phases each frame in ultrasound video capturing function at a much higher temporal resolution are essential many applications. Fulfilling these requirements particularly challenging preclinical studies involving small animals with high rates, requiring cumbersome expensive specialized hardware.We present novel method retrospective estimation directly from...
Characterizing glandular architecture in histology images of adenocarcinomas is a fundamental problem digital pathology, with important implications for computer-assisted diagnosis and grading. In this paper, we present new set features encoding the epithelium based on two recently developed vectorized persistent homology representations called persistence landscapes demonstrate their application to colorectal cancer diagnosis. On MICCAI2015 Gland Segmentation Challenge Contest dataset 165...