A5042531662- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Advanced X-ray Imaging Techniques
- Advanced MRI Techniques and Applications
- Cardiac Imaging and Diagnostics
- Radiomics and Machine Learning in Medical Imaging
- Medical Image Segmentation Techniques
- X-ray Spectroscopy and Fluorescence Analysis
- Cardiac Valve Diseases and Treatments
- Optical measurement and interference techniques
- ECG Monitoring and Analysis
- Digital Holography and Microscopy
- Medical Imaging and Analysis
- Machine Learning and Data Classification
- Seismic Imaging and Inversion Techniques
- Atomic and Subatomic Physics Research
- AI in cancer detection
- Anomaly Detection Techniques and Applications
- Brain Tumor Detection and Classification
- Image and Signal Denoising Methods
- Digital Image Processing Techniques
- Quantum Mechanics and Applications
- Experimental and Theoretical Physics Studies
- Multimodal Machine Learning Applications
- Viral Infections and Outbreaks Research
King's College London
2024
Friedrich-Alexander-Universität Erlangen-Nürnberg
2017-2022
Max Planck Institute for the Science of Light
2020-2021
Cardiovascular magnetic resonance (CMR) is an important imaging modality for the assessment and management of adult patients with congenital heart disease (CHD). However, conventional techniques three-dimensional (3D) whole-heart acquisition involve long unpredictable scan times methods that accelerate scans via k-space undersampling often rely on iterative reconstructions. Deep-learning-based reconstruction have recently attracted much interest due to their capacity provide fast...
X-ray scatter compensation is a very desirable technique in flat-panel imaging and cone-beam computed tomography. State-of-the-art U-net based removal approaches yielded promising results. However, as there are no physics' constraints applied to the output of U-Net, it cannot be ruled out that yields spurious Unfortunately, context medical imaging, those may misleading could lead wrong conclusions. To overcome this problem, we propose embed B-splines known operator into neural networks. This...
Abstract The X-ray dark-field signal can be measured with a grating-based Talbot-Lau interferometer. It measures small angle scattering of micrometer-sized oriented structures. Interestingly, the is function not only material, but also relative orientation sample, beam direction, and direction interferometer sensitivity. This property very interesting for potential tomographically reconstructing structures below imaging resolution. However, tomographic reconstruction itself substantial...
An X-ray Talbot-Lau grating interferometer enables imaging of phase contrast. It can measure material transitions that are difficult to observe with traditional attenuation X-ray, for example soft-tissue variations in medical diagnosis or weakly absorbing materials optical inspection. Unfortunately, the field view a is limited few centimeters due manufacturing constraints its gratings. For larger objects, this leads truncation projection images. In tomographic reconstruction, causes severe...
Learned iterative reconstruction algorithms for inverse problems offer the flexibility to combine analytical knowledge about problem with modules learned from data. This way, they achieve high performance while ensuring consistency measured In computed tomography, extending such approaches 2D fan-beam 3D cone-beam data is challenging due prohibitively GPU memory that would be needed train models. paper proposes use neural ordinary differential equations solve in a residual formulation via...
Unsupervised anomaly detection enables the identification of potential pathological areas by juxtaposing original images with their pseudo-healthy reconstructions generated models trained exclusively on normal images. However, clinical interpretation resultant maps presents a challenge due to lack detailed, understandable explanations. Recent advancements in language have shown capability mimicking human-like understanding and providing detailed descriptions. This raises an interesting...
Physics-inspired regularization is desired for intra-patient image registration since it can effectively capture the biomechanical characteristics of anatomical structures. However, a major challenge lies in reliance on physical parameters: Parameter estimations vary widely across literature, and properties themselves are inherently subject-specific. In this work, we introduce novel data-driven method that leverages hypernetworks to learn tissue-dependent elasticity parameters an elastic...
In this work, we introduce Progressive Growing of Patch Size, a resource-efficient implicit curriculum learning approach for dense prediction tasks. Our is defined by growing the patch size during model training, which gradually increases task's difficulty. We integrated our into nnU-Net framework and evaluated methodology on all 10 tasks Medical Segmentation Decathlon. With approach, are able to substantially reduce runtime, computational costs, CO2 emissions network training compared...
Myocardial T1 and T2 mapping play an important role in the assessment of cardiovascular disease. 3D whole-heart joint T1/T2 water/fat approaches have been recently proposed, however they require long reconstruction times. Recently a Machine learning based was proposed for motion correction corrected image undersampled free-breathing single contrast coronary MR angiography. Here, we extend this approach non-rigid motion-corrected reconstructions multi-contrast data mapping. The achieves good...
A deep learning reconstruction framework, trained in an end-to-end fashion and incorporating both a non-rigid respiratory motion estimation network motion-informed model-based network, has been previously demonstrated to enable good quality images from seven-fold undersampled acquisitions for coronary magnetic resonance angiography applications. Herein, we apply the framework whole-heart MRI scans of patients with congenital heart disease, enabling fast 7×-accelerated achieving image...
Motivation: Whole-heart CMR with high isotropic spatial resolution involves long and unpredictable scan times. Goal(s): To propose validate a super-resolution motion-corrected reconstruction framework to enable accelerated high-resolution whole-heart from lower-resolution acquisitions. Approach: Low was treated as k-space down-sampling problem, enabling the adaptation of an end-to-end iterative deep-learning network reconstruction, previously demonstrated for undersampled CMRA. Results:...
Motivation: Myocardial T1 and T2 mapping is crucial in the assessment of cardiovascular disease. 3D whole-heart joint T1/T2 approaches have been proposed, however they require long reconstruction times. Goal(s): By leveraging deep learning (DL)-based techniques, we aim to significantly reduce times for mapping, while maintaining high-quality results. Approach: Recently a group sparsity-based DL approach was proposed image undersampled multi-contrast MRI data. Here, propose extend this...
Motivation: Cardiac CINE provides dynamic images of the heart for morphology and function assessment. Single-heartbeat enables faster acquisition times study rate variations, but conventional reconstruction methods incur significant computational cost. Goal(s): This aims to speed up single-heartbeat by using deep learning reconstruction. Approach: We propose a novel, rapid, end-to-end pipeline motion estimation motion-corrected with golden-angle radial acquisition. Results: The network...