A5087146959- Medical Imaging Techniques and Applications
- Advanced X-ray and CT Imaging
- Advanced X-ray Imaging Techniques
- Photoacoustic and Ultrasonic Imaging
- Image Processing and 3D Reconstruction
- Medical Imaging and Analysis
- Medical Image Segmentation Techniques
- Cultural Heritage Materials Analysis
- Dental Radiography and Imaging
- Advanced Neural Network Applications
- Domain Adaptation and Few-Shot Learning
- X-ray Diffraction in Crystallography
- Image and Signal Denoising Methods
- Multimodal Machine Learning Applications
- Advanced Statistical Methods and Models
- Advanced MRI Techniques and Applications
- Advanced Radiotherapy Techniques
- Bayesian Modeling and Causal Inference
- Statistical Methods and Bayesian Inference
- Advanced Fluorescence Microscopy Techniques
- Computer Graphics and Visualization Techniques
- Reservoir Engineering and Simulation Methods
- Advanced Electron Microscopy Techniques and Applications
- Anatomy and Medical Technology
- Digital Radiography and Breast Imaging
Centrum Wiskunde & Informatica
2018-2024
Vitenparken
2021
Recovering a high-quality image from noisy indirect measurements is an important problem with many applications. For such inverse problems, supervised deep convolutional neural network (CNN)-based denoising methods have shown strong results, but the success of these critically depends on availability training dataset similar measurements. denoising, are available that enable without separate by assuming noise in two different pixels uncorrelated. However, this assumption does not hold for...
In order to attain anatomical models, surgical guides and implants for computer-assisted surgery, accurate segmentation of bony structures in cone-beam computed tomography (CBCT) scans is required. However, this image step often impeded by metal artifacts. Therefore, study aimed develop a mixed-scale dense convolutional neural network (MS-D network) bone CBCT affected artifacts.Training data were acquired from 20 dental An experienced medical engineer segmented the all using global...
Synchrotron X-ray tomography enables the examination of internal structure materials at submicron spatial resolution and subsecond temporal resolution. Unavoidable experimental constraints can impose dose time limits on measurements, introducing noise in reconstructed images. Convolutional neural networks (CNNs) have emerged as a powerful tool to remove from However, their training typically requires collecting dataset paired noisy high-quality which is major obstacle use practice. To...
Deep learning-based unsupervised image registration has recently been proposed, promising fast registration. However, it yet to be adopted in the online adaptive magnetic resonance imaging-guided radiotherapy (MRgRT) workflow.
Coherent X-ray microscopy is emerging as a transformative technology for neuronal imaging, with the potential to offer scalable solution reconstruction of neural circuits in millimeter sized tissue volumes. Specifically, holographic nanotomography (XNH) brings together outstanding capabilities terms contrast, spatial resolution and data acquisition speed. While recent XNH developments already enabled generating valuable datasets neurosciences, major challenge remained overcoming resolving...
Tomography is a powerful tool for reconstructing the interior of an object from series projection images. Typically, source and detector traverse standard path (e.g., circular, helical). Recently, various techniques have emerged that use more complex acquisition geometries. Current software packages require significant handwork, or lack flexibility to handle such Therefore, needed can concisely represent, visualize, compute reconstructions We present tomosipo, Python package provides these...
It is often claimed that Bayesian methods, in particular Bayes factor methods for hypothesis testing, can deal with optional stopping. We first give an overview, using elementary probability theory, of three different mathematical meanings various authors to this claim: (1) stopping rule independence, (2) posterior calibration and (3) (semi-) frequentist robustness then prove theorems the effect these claims do indeed hold a general measure-theoretic setting. For type (3), such results are...
Tomographic algorithms are often compared by evaluating them on certain benchmark datasets. For fair comparison, these datasets should ideally (i) be challenging to reconstruct, (ii) representative of typical tomographic experiments, (iii) flexible allow for different acquisition modes, and (iv) include enough samples comparison data-driven algorithms. Current approaches satisfy only some requirements, but not all. example, real-world typically a category experimental examples, restricted...
In tomography, the resolution of reconstructed 3D volume is inherently limited by pixel detector and optical phenomena. Machine learning has demonstrated powerful capabilities for super-resolution in several imaging applications. Such methods typically rely on availability high-quality training data a series similar objects. many applications existing machine cannot be used because scanning such objects either impossible or infeasible. this paper, we propose novel technique improving...
Abstract At x-ray beamlines of synchrotron light sources, the achievable time-resolution for 3D tomographic imaging interior an object has been reduced to a fraction second, enabling rapidly changing structures be examined. The associated data acquisition rates require sizable computational resources reconstruction. Therefore, full reconstruction is usually performed after scan completed. Quasi-3D reconstruction—where several interactive 2D slices are computed instead volume—has shown...
Abstract Time-resolved X-ray tomographic microscopy is an invaluable technique to investigate dynamic processes in 3D for extended time periods. Because of the limited signal-to-noise ratio caused by short exposure times and sparse angular sampling frequency, obtaining quantitative information through post-processing remains challenging requires intensive manual labor. This severely limits accessible experimental parameter space so, prevents fully exploiting capabilities dedicated...
Many of the recent successes deep learning-based approaches have been enabled by a framework flexible, composable computational blocks with their parameters adjusted through an automatic differentiation mechanism to implement various data processing tasks. In this work, we explore how same philosophy can be applied existing “classical” (i.e., non-learning) algorithms, focusing on computed tomography (CT) as application field. We apply four key design principles approach for CT workflow...
At X-ray beamlines of synchrotron light sources, the achievable time-resolution for 3D tomographic imaging interior an object has been reduced to a fraction second, enabling rapidly changing structures be examined. The associated data acquisition rates require sizable computational resources reconstruction. Therefore, full reconstruction is usually performed after scan completed. Quasi-3D -- where several interactive 2D slices are computed instead volume shown significantly more efficient,...
A major challenge in attaining customized additive manufactured (AM) skull implants is segmentation of 3D scans. Therefore, this study aimed to develop a deep learning algorithm (MSDnet) automatically segment defects computed tomography (CT) The MSDnet was trained with CT scans and corresponding virtual models patients who had undergone cranioplasty using AM implants. able unseen accurately quickly. Deep can thus remove the barriers time effort during image segmentation, thereby making more...
Neural network pruning techniques can substantially reduce the computational cost of applying convolutional neural networks (CNNs). Common methods determine which filters to remove by ranking individually, i.e., without taking into account their interdependence. In this paper, we advocate viewpoint that should consider interdependence between series consecutive operators. We propose LongEst-chAiN (LEAN) method prunes CNNs using graph-based algorithms select relevant chains convolutions. A...