A5006461848- Medical Image Segmentation Techniques
- Advanced MRI Techniques and Applications
- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Advanced Neuroimaging Techniques and Applications
- Fetal and Pediatric Neurological Disorders
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- Cardiac Imaging and Diagnostics
- Functional Brain Connectivity Studies
- Neonatal and fetal brain pathology
- Medical Imaging and Analysis
- AI in cancer detection
- Brain Tumor Detection and Classification
- COVID-19 diagnosis using AI
- Cardiovascular Function and Risk Factors
- Privacy-Preserving Technologies in Data
- Traumatic Brain Injury and Neurovascular Disturbances
- Retinal Imaging and Analysis
- Dementia and Cognitive Impairment Research
- Advanced X-ray and CT Imaging
- Machine Learning in Healthcare
- Image Retrieval and Classification Techniques
- Cardiac Valve Diseases and Treatments
- Advanced Vision and Imaging
Technical University of Munich
2020-2025
Institute of Group Analysis
2016-2025
Imperial College London
2016-2025
Klinikum rechts der Isar
2020-2025
Munich Center for Machine Learning
2024-2025
Institute for Sports Medicine
2024
University of Cambridge
2024
University of Regensburg
2024
Fulda University of Applied Sciences
2024
Analysis Group (United States)
2023
Recently, several models based on deep neural networks have achieved great success in terms of both reconstruction accuracy and computational performance for single image super-resolution. In these methods, the low resolution (LR) input is upscaled to high (HR) space using a filter, commonly bicubic interpolation, before reconstruction. This means that super-resolution (SR) operation performed HR space. We demonstrate this sub-optimal adds complexity. paper, we present first convolutional...
In this paper the authors present a new approach for nonrigid registration of contrast-enhanced breast MRI. A hierarchical transformation model motion has been developed. The global is modeled by an affine while local described free-form deformation (FFD) based on B-splines. Normalized mutual information used as voxel-based similarity measure which insensitive to intensity changes result contrast enhancement. Registration achieved minimizing cost function, represents combination associated...
We propose a novel attention gate (AG) model for medical imaging that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn suppress irrelevant regions in an input image while highlighting salient features useful specific task. This enables us eliminate the necessity using explicit external tissue/organ localisation modules cascaded convolutional neural networks (CNNs). can be easily integrated into standard CNN architectures...
We propose a dual pathway, 11-layers deep, three-dimensional Convolutional Neural Network for the challenging task of brain lesion segmentation. The devised architecture is result an in-depth analysis limitations current networks proposed similar applications. To overcome computational burden processing 3D medical scans, we have efficient and effective dense training scheme which joins adjacent image patches into one pass through network while automatically adapting to inherent class...
We propose a novel attention gate (AG) model for medical image analysis that automatically learns to focus on target structures of varying shapes and sizes. Models trained with AGs implicitly learn suppress irrelevant regions in an input while highlighting salient features useful specific task. This enables us eliminate the necessity using explicit external tissue/organ localisation modules when convolutional neural networks (CNNs). can be easily integrated into standard CNN models such as...
Inspired by recent advances in deep learning, we propose a framework for reconstructing dynamic sequences of 2-D cardiac magnetic resonance (MR) images from undersampled data using cascade convolutional neural networks (CNNs) to accelerate the acquisition process. In particular, address case where are acquired aggressive Cartesian undersampling. First, show that when each image frame is reconstructed independently, proposed method outperforms state-of-the-art compressed sensing approaches,...
Incorporation of prior knowledge about organ shape and location is key to improve performance image analysis approaches. In particular, priors can be useful in cases where images are corrupted contain artefacts due limitations acquisition. The highly constrained nature anatomical objects well captured with learning based techniques. However, most recent promising techniques such as CNN segmentation it not obvious how incorporate knowledge. State-of-the-art methods operate pixel-wise...
Cardiovascular resonance (CMR) imaging is a standard modality for assessing cardiovascular diseases (CVDs), the leading cause of death globally. CMR enables accurate quantification cardiac chamber volume, ejection fraction and myocardial mass, providing information diagnosis monitoring CVDs. However, years, clinicians have been relying on manual approaches image analysis, which time consuming prone to subjective errors. It major clinical challenge automatically derive quantitative clinically...
Accelerating the data acquisition of dynamic magnetic resonance imaging leads to a challenging ill-posed inverse problem, which has received great interest from both signal processing and machine learning communities over last decades. The key ingredient problem is how exploit temporal correlations MR sequence resolve aliasing artifacts. Traditionally, such observation led formulation an optimization was solved using iterative algorithms. Recently, however, deep learning-based approaches...
In this paper, we show how the concept of statistical deformation models (SDMs) can be used for construction average anatomy and their variability. SDMs are built by performing a analysis deformations required to map anatomical features in one subject into corresponding another subject. The is similar shape (SSMs) which capture information about shapes across population, but offers several advantages over SSMs. First, constructed directly from images such as three-dimensional (3-D) magnetic...
In this paper, we propose DeepCut, a method to obtain pixelwise object segmentations given an image dataset labelled weak annotations, in our case bounding boxes. It extends the approach of well-known GrabCut[1] include machine learning by training neural network classifier from box annotations. We formulate problem as energy minimisation over densely-connected conditional random field and iteratively update targets segmentations. Additionally, variants DeepCut compare those naïve CNN under...