A5033880300- Sparse and Compressive Sensing Techniques
- Medical Imaging Techniques and Applications
- Medical Image Segmentation Techniques
- Image and Signal Denoising Methods
- Radiomics and Machine Learning in Medical Imaging
- Numerical methods in inverse problems
- Generative Adversarial Networks and Image Synthesis
- AI in cancer detection
- Advanced X-ray and CT Imaging
- Advanced MRI Techniques and Applications
- Advanced Neural Network Applications
- Photoacoustic and Ultrasonic Imaging
- Neural Networks and Applications
- Model Reduction and Neural Networks
- Advanced Image Processing Techniques
- COVID-19 diagnosis using AI
- Stochastic Gradient Optimization Techniques
- Cell Image Analysis Techniques
- Advanced Numerical Analysis Techniques
- Machine Learning in Healthcare
- Brain Tumor Detection and Classification
- Advanced Image Fusion Techniques
- Advanced Neuroimaging Techniques and Applications
- Advanced Mathematical Modeling in Engineering
- Artificial Intelligence in Healthcare and Education
University of Cambridge
2016-2025
Engineering and Physical Sciences Research Council
2021-2024
Bridge University
2024
University of Bath
2021-2024
Norwegian University of Science and Technology
2024
University of Bonn
2023
Hausdorff Center for Mathematics
2023
Maxwell Institute for Mathematical Sciences
2023
Heriot-Watt University
2023
Delft University of Technology
2023
Machine learning methods offer great promise for fast and accurate detection prognostication of COVID-19 from standard-of-care chest radiographs (CXR) computed tomography (CT) images. Many articles have been published in 2020 describing new machine learning-based models both these tasks, but it is unclear which are potential clinical utility. In this systematic review, we search EMBASE via OVID, MEDLINE PubMed, bioRxiv, medRxiv arXiv papers preprints uploaded January 1, to October 3,...
Recent research in inverse problems seeks to develop a mathematically coherent foundation for combining data-driven models, and particular those based on deep learning, with domain-specific knowledge contained physical–analytical models. The focus is solving ill-posed that are at the core of many challenging applications natural sciences, medicine life as well engineering industrial applications. This survey paper aims give an account some main contributions problems.
Automatic segmentation methods are an important advancement in medical image analysis. Machine learning techniques, and deep neural networks particular, the state-of-the-art for most tasks. Issues with class imbalance pose a significant challenge datasets, lesions often occupying considerably smaller volume relative to background. Loss functions used training of algorithms differ their robustness imbalance, direct consequences model convergence. The commonly loss based on either cross...
In the literature of remote sensing, deep models with multiple layers have demonstrated their potentials in learning abstract and invariant features for better representation classification hyperspectral images. The usual supervised models, such as convolutional neural networks, need a large number labeled training samples to learn model parameters. However, real-world image task provides only limited samples. This paper adopts another popular model, i.e., belief networks (DBNs), deal this...
Colonoscopy remains the gold-standard screening for colorectal cancer. However, significant miss rates polyps have been reported, particularly when there are multiple small adenomas. This presents an opportunity to leverage computer-aided systems support clinicians and reduce number of missed.In this work we introduce Focus U-Net, a novel dual attention-gated deep neural network, which combines efficient spatial channel-based attention into single Gate module encourage selective learning...
Removing the bias and variance of multicentre data has always been a challenge in large scale digital healthcare studies, which requires ability to integrate clinical features extracted from acquired by different scanners protocols improve stability robustness. Previous studies have described various computational approaches fuse single modality datasets. However, these surveys rarely focused on evaluation metrics lacked checklist for harmonisation studies. In this systematic review, we...
Compressed sensing (CS) has been playing a key role in accelerating the magnetic resonance imaging (MRI) acquisition process. With resurgence of artificial intelligence, deep neural networks and CS algorithms are being integrated to redefine state art fast MRI. The past several years have witnessed substantial growth complexity, diversity, performance deep-learning-based techniques that dedicated In this meta-analysis, we systematically review for MRI, describe model designs, highlight...
Abstract Partial differential equations (PDEs) play a fundamental role in the mathematical modelling of many processes and systems physical, biological other sciences. To simulate such systems, solutions PDEs often need to be approximated numerically. The finite element method, for instance, is usual standard methodology do so. recent success deep neural networks at various approximation tasks has motivated their use numerical solution PDEs. These so-called physics-informed variants have...
The Cahn–Hilliard equation is a nonlinear fourth order diffusion originating in material science for modeling phase separation and coarsening binary alloys. inpainting of images using the new approach image processing. In this paper we discuss stationary state proposed model introduce generalization grayvalue bounded variation. This realized by subgradients total variation functional within flow, which leads to structure with smooth curvature level sets.
We propose a stochastic extension of the primal-dual hybrid gradient algorithm studied by Chambolle and Pock in 2011 to solve saddle point problems that are separable dual variable. The analysis is carried out for general convex-concave either partially smooth, strongly convex or fully convex. perform arbitrary samplings variables, we obtain known deterministic results as special case. Several variants our method significantly outperform variant on variety imaging tasks.
Higher order equations, when applied to image inpainting, have certain advantages over second such as continuation of both edge and intensity information larger distances.Discretizing a fourth evolution equation with brute force method may restrict the time steps size up ∆x 4 where denotes step spatial grid.In this work we present efficient semi-implicit schemes that are guaranteed be unconditionally stable.We explain main idea these applications in processing for inpainting Cahn-Hilliard...
We consider a bilevel optimisation approach for parameter learning in higher-order total variation image reconstruction models. Apart from the least squares cost functional, naturally used learning, we propose and analyse an alternative based on Huber-regularised TV seminorm. Differentiability properties of solution operator are verified first-order optimality system is derived. Based adjoint information, combined quasi-Newton/semismooth Newton algorithm proposed numerical problems....
We propose a nonsmooth PDE-constrained optimization approach for the determination of correctnoise model in total variation (TV) image denoising. Anoptimization problem weightscorresponding to different types noise distributions is stated and existence an optimal solution isproved. A tailored regularization approximation optimalparameter values proposed thereafter its consistencystudied. Additionally, differentiability operatoris proved optimality system characterizing optimalsolutions each...
Inverse Problems in medical imaging and computer vision are traditionally solved using purely model-based methods. Among those variational regularization models one of the most popular approaches. We propose a new framework for applying data-driven approaches to inverse problems, neural network as functional. The learns discriminate between distribution ground truth images unregularized reconstructions. Once trained, is applied problem by solving corresponding problem. Unlike other...
The liquid and glass states of metal-organic frameworks (MOFs) have recently become interest due to the potential for liquid-phase separations ion transport, alongside fundamental nature latter as a new, fourth category melt-quenched glass. Here we show that MOF state can be blended with another component, resulting in domain structured single, tailorable transition. Intra-domain connectivity short range order is confirmed by nuclear magnetic resonance spectroscopy pair distribution function...
A central problem in hyperspectral image classification is obtaining high accuracy when using a limited amount of labelled data. In this paper we present novel graph-based framework, which aims to tackle the presence large scale data input. Our approach utilises superpixel method, specifically designed for data, define meaningful local regions an image, with probability share same label. We then extract spectral and spatial features from these use produce contracted weighted...
Recent studies on the adversarial vulnerability of neural networks have shown that models trained to be more robust attacks exhibit interpretable saliency maps than their non-robust counterparts. We aim quantify this behavior by considering alignment between input image and map. hypothesize as distance decision boundary grows,so does alignment. This connection is strictly true in case linear models. confirm these theoretical findings with experiments based a local Lipschitz regularization...
Score-based diffusion models have emerged as one of the most promising frameworks for deep generative modelling. In this work we conduct a systematic comparison and theoretical analysis different approaches to learning conditional probability distributions with score-based models. particular, prove results which provide justification successful estimators score. Moreover, introduce multi-speed framework, leads new estimator score, performing on par previous state-of-the-art approaches. Our...
The “fast iterative shrinkage-thresholding algorithm,” a.k.a. FISTA, is one of the most well known first-order optimization scheme in literature, as it achieves worst-case $O(1/k^2)$ optimal convergence rate for objective function value. However, despite such an theoretical rate, practice (local) oscillatory behavior FISTA often damps its efficiency. Over past years, various efforts have been made literature to improve practical performance monotone restarting and backtracking strategies. In...
Abstract Cell Painting is a high-content image-based assay applied in drug discovery to predict bioactivity, assess toxicity and understand mechanisms of action chemical genetic perturbations. We investigate label-free by predicting the five fluorescent channels from brightfield input. train validate two deep learning models with dataset representing 17 batches, we evaluate on batches treated compounds phenotypic set. The mean Pearson correlation coefficient predicted images across all 0.84....
Semantic segmentation has been widely investigated in the community, which state-of-the-art techniques are based on supervised models. Those models have reported unprecedented performance at cost of requiring a large set high quality masks for training. Obtaining such annotations is highly expensive and time consuming, particular, semantic where pixel-level required. In this work, we address problem by proposing holistic solution framed as self-training framework semi-supervised...