- AI in cancer detection
- Image and Signal Denoising Methods
- Sparse and Compressive Sensing Techniques
- Non-Invasive Vital Sign Monitoring
- Cardiac Imaging and Diagnostics
- Digital Imaging for Blood Diseases
- Radiomics and Machine Learning in Medical Imaging
- Blind Source Separation Techniques
- Medical Image Segmentation Techniques
- Neonatal Respiratory Health Research
- Cardiac Arrest and Resuscitation
- Advanced MRI Techniques and Applications
- Colorectal Cancer Screening and Detection
- Neonatal and fetal brain pathology
- Cell Image Analysis Techniques
- Healthcare Technology and Patient Monitoring
- Brain Tumor Detection and Classification
- Image Retrieval and Classification Techniques
- Heart Rate Variability and Autonomic Control
- Acute Ischemic Stroke Management
- Coronary Interventions and Diagnostics
- Retinal Imaging and Analysis
- Advanced X-ray and CT Imaging
- Advanced Adaptive Filtering Techniques
- ECG Monitoring and Analysis
University of Stavanger
2016-2025
Erasmus MC Cancer Institute
2024
University Medical Center
2024
Stavanger University Hospital
2024
INCLIVA Health Research Institute
2024
Universitat de València
2024
Biomedical Research Institute
2024
Cardiovascular Institute of the South
2018
Antwerp University Hospital
2018
University of Antwerp
2018
We address the problem of finding sparse solutions to an underdetermined system equations when there are multiple measurement vectors having same, but unknown, sparsity structure. The single solution has been extensively studied in past. Although known be NP-hard, many single-measurement suboptimal algorithms have formulated that found utility different applications. Here, we consider depth extension two classes algorithms-Matching Pursuit (MP) and FOCal Underdetermined System Solver...
A frame design technique for use with vector selection algorithms, example matching pursuits (MP), is presented. The algorithm iterative and requires a training set of signal vectors. algorithm, called method optimal directions (MOD), an improvement the presented by Engan, Aase Husoy see (Proc. ICASSP '98, Seattle, USA, p.1817-20, 1998). MOD applied to speech electrocardiogram (ECG) signals, designed frames are tested on signals outside sets. Experiments demonstrate that approximation...
Algorithms for data-driven learning of domain-specific overcomplete dictionaries are developed to obtain maximum likelihood and a posteriori dictionary estimates based on the use Bayesian models with concave/Schur-concave (CSC) negative log priors. Such priors appropriate obtaining sparse representations environmental signals within an appropriately chosen (environmentally matched) dictionary. The elements can be interpreted as concepts, features, or words capable succinct expression events...
We present the recursive least squares dictionary learning algorithm, RLS-DLA, which can be used for overcomplete dictionaries sparse signal representation. Most DLAs presented earlier, example ILS-DLA and K-SVD, update after a batch of training vectors has been processed, usually using whole set as one batch. The is iteratively to gradually improve dictionary. approach in RLS-DLA continuous each vector being processed. core algorithm compact effectively implemented. derived very much along...
We develop robust methods for subset selection based on the minimization of diversity measures. A Bayesian framework is used to account noise in data and a maximum posteriori (MAP) estimation procedure leads an iterative which regularized version focal underdetermined system solver (FOCUSS) algorithm. The convergence FOCUSS algorithm established it shown that stable fixed points are sparse. investigate three different criteria choosing regularization parameter: quality fit; sparsity...
Abstract Accurate observations at birth and during newborn resuscitation are fundamental for quality improvement initiatives research. However, manual data collection methods often lack consistency objectivity, not scalable, may raise privacy concerns. The NewbornTime project aims to develop an AI system that generates accurate timelines from events by automated video recording processing, providing a source of objective consistent data. This work describe the implementation is necessary...
The method of optimal directions (MOD) is an iterative for designing frames sparse representation purposes using a training set. In this paper we use designed by MOD in multiframe compression (MFC) scheme. Both the and MFC need vector selection algorithm, orthogonal matching pursuit (OMP) used paper. scheme several different are used, each optimized fixed number selected frame vectors approximation. We apply to ECG signals, do experiments with both size variable on Compared traditional...
The recently presented recursive least squares dictionary learning algorithm (RLS-DLA) is tested in a general image compression application. Dictionaries are learned the pixel domain and 9/7 wavelet domain, then straightforward scheme. Results compared with state-of-the-art methods. proposed scheme using RLS DLA dictionaries per forms better than by other rate just below JPEG 2000 which promising considering simple entropy coding used.
This paper investigates discrimination capabilities in the texture of fundus images to differentiate between pathological and healthy images. For this purpose, performance local binary patterns (LBP) as a descriptor for retinal has been explored compared with other descriptors such LBP filtering phase quantization. The goal is distinguish diabetic retinopathy (DR), age-related macular degeneration (AMD), normal analyzing retina background avoiding previous lesion segmentation stage. Five...
There is a need to monitor the heart rhythm in resuscitation improve treatment quality. Resuscitation rhythms are categorized into: ventricular tachycardia (VT), fibrillation (VF), pulseless electrical activity (PEA), asystole (AS), and pulse-generating (PR). Manual annotation of time-consuming infeasible for large datasets. Our objective was develop ECG-based algorithms retrospective automatic classification cardiac rhythms.The dataset consisted 1631 3-s ECG segments with clinical...
Aim: Our aim was to automatically estimate the blood velocity in coronary arteries using cine X-ray angiographic sequence. Estimating is a key approach investigating patients with angina pectoris and no significant artery disease. Blood estimation central assessing flow reserve. Methods & Results: A multi-step automatic method for based on information extracted solely from angiography sequence obtained by invasive selective catheterization developed. The includes 1) an iterative process of...
Dementia is an evolving challenge in society, and no disease-modifying treatment exists. Diagnosis can be demanding MR imaging may aid as a noninvasive method to increase prediction accuracy. We explored the use of 2D local binary pattern (LBP) extracted from FLAIR T1 images brain combined with Random Forest classifier attempt discern patients Alzheimer's disease (AD), Lewy body dementia (LBD), normal controls (NC). Analysis was conducted areas white matter lesions (WML) all (WM). Results...
Modern cancer diagnostics involves extracting tissue specimens from suspicious areas and conducting histotechnical procedures to prepare a digitized glass slide, called Whole Slide Image (WSI), for further examination. These frequently introduce different types of artifacts in the obtained WSI, histological might influence Computational Pathology (CPATH) systems down diagnostic pipeline if not excluded or handled. Deep Convolutional Neural Networks (DCNNs) have achieved promising results...
In pathology labs worldwide, we see an increasing number of tissue samples that need to be assessed without the same increase in pathologists. Computational pathology, where digital scans histological called whole-slide images (WSI) are processed by computational tools, can help for pathologists and is gaining research interests. Most effort has been given classify slides as being cancerous or not, localization regions, “big-four” cancer: breast, lung, prostate, bowel. Urothelial carcinoma,...
Bladder cancer patients' stratification into risk groups relies on grade, stage and clinical factors. For non-muscle invasive bladder cancer, T1 tumours that invade the subepithelial tissue are high-risk lesions with a high probability to progress an aggressive muscle-invasive disease. Detecting cancerous areas is main factor for dictating treatment strategy patient. However, defining invasion often subject intra/interobserver variability among pathologists, thus leading over or...
Approximately 12 percent of men will suffer from prostate cancer during their lifetime. While some types this grow slowly and need minimal treatment, others are aggressive spread quickly. In order to be able provide a fast accurate diagnosis, computer vision techniques widely used help pathologists in the diagnosis prognosis tasks. work, new Content-Based Medical Image Retrieval (CBMIR) method is presented. The method, which named ResCAE, presents modified Convolutional Auto-Encoder (CAE)...
Histological tissue examination has been a longstanding practice for cancer diagnosis where pathologists identify the presence of tumors on glass slides. Slides acquired from laboratory routine may contain unintentional artifacts due to complications in surgical resection. Blood and damaged are two common problems associated with transurethral resection bladder tumor. Differences histotechnical procedures among laboratories also result color variations minor inconsistencies outcome. A...