A5036336660- Acute Ischemic Stroke Management
- Topic Modeling
- Medical Imaging Techniques and Applications
- Medical Image Segmentation Techniques
- Intracerebral and Subarachnoid Hemorrhage Research
- Traumatic Brain Injury and Neurovascular Disturbances
- Cerebrovascular and Carotid Artery Diseases
- Radiomics and Machine Learning in Medical Imaging
- Natural Language Processing Techniques
- Authorship Attribution and Profiling
- Speech Recognition and Synthesis
- Medical Imaging and Analysis
- Oral Health Pathology and Treatment
- Cardiovascular Disease and Adiposity
- Brain Tumor Detection and Classification
- Advanced MRI Techniques and Applications
- Cardiac Imaging and Diagnostics
- Psoriasis: Treatment and Pathogenesis
- Machine Learning in Healthcare
- Advanced Neural Network Applications
- Smart Agriculture and AI
- Multi-Agent Systems and Negotiation
- Advanced X-ray and CT Imaging
- Functional Brain Connectivity Studies
- Retinal Imaging and Analysis
Radboud University Nijmegen
2016-2021
Radboud University Medical Center
2016-2021
California University of Pennsylvania
2021
University of Colorado System
2021
University Medical Center
2016-2019
National Hospital for Neurology and Neurosurgery
2008
University College London
2008
Current hardware restrictions pose limitations on the use of convolutional neural networks for medical image analysis. There is a large trade-off between network architecture and input size. For this reason, identification classification tasks are commonly approached with patch or region-based methods often utilizing only local contextual information during training at inference. Here, method presented intracranial hemorrhage (ICH) in three-dimensional (3D) non-contrast computed tomography...
A 3-dimensional (3D) convolutional neural network is presented for the segmentation and quantification of spontaneous intracerebral haemorrhage (ICH) in non-contrast computed tomography (NCCT). The method utilises a combination contextual information on multiple scales fast fully automatic dense predictions. To handle large class imbalance present data, weight map introduced during training. was evaluated two datasets 25 50 patients respectively. reference standard consisted manual...
The Psoriasis Area and Severity Index (PASI) score is commonly used in clinical practice research to monitor disease severity determine treatment efficacy. Automating the PASI with deep learning algorithms, like Convolutional Neural Networks (CNNs), could enable objective efficient scoring.To assess performance of image-based automated scoring anatomical regions by CNNs compare physicians.Imaging series were matched subscores determined real life treating physician. trained using...
A robust method is presented for the segmentation of full cerebral vasculature in 4-dimensional (4D) computed tomography (CT). The consists candidate vessel selection, feature extraction, random forest classification and postprocessing. Image features include among others weighted temporal variance image parameters, including entropy, an intensity histogram a local region at different scales. These parameters revealed to be strong detection vessels regardless shape size. was trained tested...
4D CT imaging has a great potential for use in stroke workup. A fully convolutional neural network (CNN) 3D multiclass segmentation is presented, which can be trained end-to-end from sparse 2D annotations. The CNN was and validated on 42 acquisitions of the brain patients with suspicion acute ischemic stroke. White matter, gray cerebrospinal fluid, vessels were annotated by two observers. mean Dice coefficients, contour distances, absolute volume differences were, respectively, 0.87 ± 0.04,...
To implement and test a deep learning approach for the segmentation of arterial venous cerebral vasculature with four-dimensional (4D) CT angiography.Patients who had undergone 4D angiography suspicion acute ischemic stroke were retrospectively identified. A total 390 patients evaluated in 2014 (n = 113) or 2018 277) included this study, each patient having one angiographic scan. One hundred from randomly selected, arteries veins on their scans manually annotated by five experienced...
Segmentation of anatomical structures is fundamental in the development computer aided diagnosis systems for cerebral pathologies. Manual annotations are laborious, time consuming and subject to human error observer variability. Accurate quantification cerebrospinal fluid (CSF) can be employed as a morphometric measure patient outcome prediction. However, segmenting CSF non-contrast CT images complicated by low soft tissue contrast image noise. In this paper we propose state-of-the-art...
The assessment of the presence intracranial hemorrhage is a crucial step in work-up patients requiring emergency care. Fast and accurate detection can aid treating physicians by not only expediting guiding diagnosis, but also supporting choices for secondary imaging, treatment intervention. However, automatic complicated variation appearance on non-contrast CT images as result differences etiology location. We propose method using convolutional neural network (CNN) hemorrhage. trained...
Training models to act as agents that can effectively navigate and perform actions in a complex environment, such web browser, has typically been challenging due lack of training data. Large language (LLMs) have recently demonstrated some capability novel environments zero-shot or few-shot fashion, purely guided by natural instructions prompts. Recent research also LLMs the exceed their base performance through self-improvement, i.e. fine-tuning on data generated model itself. In this work,...
The goal of text style transfer is to transform the texts while preserving their original meaning, often with only a few examples target style. Existing methods generally rely on few-shot capabilities large language models or complex controllable generation approaches that are inefficient and underperform fluency metrics. We introduce TinyStyler, lightweight but effective approach, which leverages small model (800M params) pre-trained authorship embeddings perform efficient, transfer....
Rebecca Iglesias-Flores, Megha Mishra, Ajay Patel, Akanksha Malhotra, Reno Kriz, Martha Palmer, Chris Callison-Burch. Proceedings of the Second Workshop on Data Science with Human in Loop: Language Advances. 2021.