A5055308236- COVID-19 Clinical Research Studies
- Platelet Disorders and Treatments
- Inflammatory Biomarkers in Disease Prognosis
- Digital Imaging for Blood Diseases
- Cell Image Analysis Techniques
- AI in cancer detection
- Retinal Imaging and Analysis
- COVID-19 diagnosis using AI
- COVID-19 and healthcare impacts
- Anomaly Detection Techniques and Applications
- Long-Term Effects of COVID-19
- Face and Expression Recognition
- Image and Object Detection Techniques
- Medical Image Segmentation Techniques
- Medical Imaging Techniques and Applications
- Face recognition and analysis
- Domain Adaptation and Few-Shot Learning
- Osteoarthritis Treatment and Mechanisms
- Blind Source Separation Techniques
- ECG Monitoring and Analysis
- Advanced Neural Network Applications
- Time Series Analysis and Forecasting
- Gene expression and cancer classification
- Adversarial Robustness in Machine Learning
- Music and Audio Processing
University of Virginia
2019-2025
North South University
2024-2025
National Institutes of Health Clinical Center
2024
University of North Carolina at Chapel Hill
2024
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...
Significance Many diseases lack visual cues in early stages, eluding image-based detection. Today, osteoarthritis is diagnosed at an irreversible stage after bone damage has occurred. This research demonstrates that detection healthy individuals may be possible 3 y prior to symptoms or damage. We describe a technique combining mass transport theory with statistical pattern recognition, discovering sensitive cartilage phenotypes predict future progression high accuracy. As our approach based...
This paper presents a new end-to-end signal classification method using the signed cumulative distribution transform (SCDT). We adopt transport generative model to define problem. then make use of mathematical properties SCDT render problem easier in domain, and solve for class an unknown sample nearest local subspace (NLS) search algorithm domain. Experiments show that proposed provides high accuracy results while being computationally cheap, data efficient, robust out-of-distribution...
ABSTRACT Alterations in nuclear morphology are useful adjuncts and even diagnostic tools used by pathologists the diagnosis grading of many tumors, particularly malignant tumors. Large datasets such as TCGA Human Protein Atlas, combination with emerging machine learning statistical modeling methods, feature extraction deep techniques, can be to extract meaningful knowledge from images nuclei, cancerous Here, we describe a new technique based on mathematics optimal transport for information...
Microvascular thrombosis is a typical symptom of COVID-19 and shows similarities to thrombosis. Using microfluidic imaging flow cytometer, we measured the blood 181 samples 101 non-COVID-19 samples, resulting in total 6.3 million bright-field images. We trained convolutional neural network distinguish single platelets, platelet aggregates, white cells performed classical image analysis for each subpopulation individually. Based on derived single-cell features population, machine learning...
This paper presents a new method to classify 1D signals using the signed cumulative distribution transform (SCDT). The proposed exploits certain linearization properties of SCDT render problem easier solve in space. uses nearest subspace search technique domain provide non-iterative, effective, and simple implement classification algorithm. Experiments show that outperforms state-of-the-art neural networks very low number training samples is also robust out-of-distribution examples on...
Abstract Learning from point sets is an essential component in many computer vision and machine learning applications. Native, unordered, permutation invariant set structure space challenging to model, particularly for classification under spatial deformations. Here we propose a framework classifying experiencing certain types of deformations, with particular emphasis on datasets featuring affine Our approach employs the Linear Optimal Transport (LOT) transform obtain linear embedding...
We present a new method for face recognition from digital images acquired under varying illumination conditions. The is based on mathematical modeling of local gradient distributions using the Radon Cumulative Distribution Transform (R-CDT). demonstrate that lighting variations cause certain types deformations image which, when expressed in R-CDT domain, can be modeled as subspace. Face then performed nearest subspace domain distributions. Experiment results proposed outperforms other...
ABSTRACT 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...
Here we describe a new image representation technique based on the mathematics of transport and optimal transport. The method relies combination well-known Radon transform for images recent signal called Signed Cumulative Distribution Transform. newly proposed generalizes previous transport-related methods to arbitrary functions (images), thus can be used in more applications. We transform, some its mathematical properties demonstrate ability partition classes with real simulated data. In...
Cell image classification methods are currently being used in numerous applications cell biology and medicine. Applications include understanding the effects of genes drugs screening experiments, role subcellular localization different proteins, as well diagnosis prognosis cancer from images acquired using cytological histological techniques. We review three approaches for classification: numerical feature extraction, end to with neural networks, transport-based morphometry. In addition, we...
We present a new supervised image classification method applicable to broad class of deformation models. The makes use the previously described Radon Cumulative Distribution Transform (R-CDT) for data, whose mathematical properties are exploited express data in form that is more suitable machine learning. While certain operations such as translation, scaling, and higher-order transformations challenging model native space, we show R-CDT can capture some these variations thus render...
ABSTRACT Microvascular thrombosis is a typical symptom of COVID-19 and shows similarities to thrombosis. Using microfluidic imaging flow cytometer, we measured the blood 181 samples 101 non-COVID-19 samples, resulting in total 6.3 million bright-field images. We trained convolutional neural network distinguish single platelets, platelet aggregates, white cells performed classical image analysis for each subpopulation individually. Based on derived single-cell features population, machine...
Abstract A characteristic clinical feature of COVID-19 is the frequent occurrence thrombotic events. Furthermore, many cases multiorgan failure are in nature. Since outbreak COVID-19, D-dimer testing has been used extensively to evaluate COVID-19-associated thrombosis, but does not provide a complete view disease because it probes blood coagulation, platelet activity. Due this limitation, fails account for events which occur despite low levels, such as sudden stroke young patients and...
This paper presents a new method to classify 1D signals using the signed cumulative distribution transform (SCDT). The proposed exploits certain linearization properties of SCDT render problem easier solve in space. uses nearest subspace search technique domain provide non-iterative, effective, and simple implement classification algorithm. Experiments show that outperforms state-of-the-art neural networks very low number training samples is also robust out-of-distribution examples on...