Real-time segmentation for tomographic imaging

Schoonhoven, Richard; Buurlage, Jan‐Willem; Pelt, Daniël M.; Batenburg, Kees Joost

doi:10.1109/mlsp49062.2020.9231642

Cited by 3 publications

(3 citation statements)

References 28 publications

(41 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…With the automatic lens changing mechanism we can easily pad high resolution with low resolution projections and correct the high-resolution scans from local tomography artifacts (Xiao et al, 2007). Finally, we plan to add AI-based methods (Schoonhoven et al, 2020;Tekawade et al, 2021) for detecting events and trigger data saving automatically. In fast-evolving dynamic systems, automatic segmentation, classification, and detection may allow for steering tomographic experiments, e.g.…”

Section: Discussionmentioning

confidence: 99%

Real-time streaming tomographic reconstruction with on-demand data capturing and 3D zooming to regions of interest

Никитин¹,

Tekawade²,

Duchkov³

et al. 2022

J Synchrotron Radiat

View full text Add to dashboard Cite

Complex dynamic tomographic experiments at brilliant X-ray light sources require real-time feedback on the sample changes with respect to environmental conditions, selecting representative regions of interest for high-resolution scanning, and on-demand data saving mechanisms for storing only relevant projections acquired by fast area detectors and reducing data volumes. Here the implementation details of a 3D real-time imaging monitoring instrument, with zooming to a volume of interest with easy-to-use visualization via ImageJ, a tool familiar to most beamline users, is presented. The instrument relies on optimized data flow between the detector and processing machines and is implemented on commodity computers. The instrument has been developed at beamline 2-BM of the Advanced Photon Source, where the automatic lens changing mechanism for zooming is implemented with an Optique Peter microscope. Performance tests demonstrate the ability to process more than 3 GB of projection data per second and generate real-time 3D zooming with different magnification. These new capabilities are essential for new APS Upgrade instruments such as the projection microscope under development at beamline 32-ID. The efficacy of the proposed instrument was demonstrated during an in situ tomographic experiment on ice and gas hydrate formation in porous samples.

show abstract

Section: Discussionmentioning

confidence: 99%

Real-time streaming tomographic reconstruction with on-demand data capturing and 3D zooming to regions of interest

Никитин¹,

Tekawade²,

Duchkov³

et al. 2022

J Synchrotron Radiat

View full text Add to dashboard Cite

show abstract

“…The bubbles are to be segmented and separated from the noise and tomographic artifacts. The dataset is described in detail in [42]. The dataset consists of 512 × 512 images, split into 9216 training images, 2048 validation images, and 1536 test images.…”

Section: Real-world Dynamic Ct Datasetmentioning

confidence: 99%

“…Therefore, we use the F1- score (Dice coefficient). For our experiments we use the same MS-D architecture as used in [42]. The MS-D network that was used as a starting point for pruning achieved an F1-score of 0.8816.…”

Section: Real-world Dynamic Ct Datasetmentioning

confidence: 99%

LEAN: graph-based pruning for convolutional neural networks by extracting longest chains

Schoonhoven¹,

Hendriksen²,

Pelt³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Convolutional neural networks (CNNs) have proven to be highly successful at a range of image-to-image tasks. CNNs can be computationally expensive, which can limit their applicability in practice. Model pruning can improve computational efficiency by sparsifying trained networks. Common methods for pruning CNNs determine what convolutional filters to remove by ranking filters on an individual basis. However, filters are not independent, as CNNs consist of chains of convolutions, which can result in suboptimal filter selection.We propose a novel pruning method, LongEst-chAiN (LEAN) pruning, which takes the interdependency between the convolution operations into account. We propose to prune CNNs by using graph-based algorithms to select relevant chains of convolutions. A CNN is interpreted as a graph, with the operator norm of each convolution as distance metric for the edges. LEAN pruning iteratively extracts the highest value path from the graph to keep. In our experiments, we test LEAN pruning for several imageto-image tasks, including the well-known CamVid dataset. LEAN pruning enables us to keep just 0.5%-2% of the convolutions without significant loss of accuracy. When pruning CNNs with LEAN, we achieve a higher accuracy than pruning filters individually, and different pruned substructures emerge.

show abstract

Just-in-time deep learning for real-time X-ray computed tomography

Graas,

Coban,

Batenburg

et al. 2023

Sci Rep

View full text Add to dashboard Cite

Real-time X-ray tomography pipelines, such as implemented by RECAST3D, compute and visualize tomographic reconstructions in milliseconds, and enable the observation of dynamic experiments in synchrotron beamlines and laboratory scanners. For extending real-time reconstruction by image processing and analysis components, Deep Neural Networks (DNNs) are a promising technology, due to their strong performance and much faster run-times compared to conventional algorithms. DNNs may prevent experiment repetition by simplifying real-time steering and optimization of the ongoing experiment. The main challenge of integrating DNNs into real-time tomography pipelines, however, is that they need to learn their task from representative data before the start of the experiment. In scientific environments, such training data may not exist, and other uncertain and variable factors, such as the set-up configuration, reconstruction parameters, or user interaction, cannot easily be anticipated beforehand, either. To overcome these problems, we developed just-in-time learning, an online DNN training strategy that takes advantage of the spatio-temporal continuity of consecutive reconstructions in the tomographic pipeline. This allows training and deploying comparatively small DNNs during the experiment. We provide software implementations, and study the feasibility and challenges of the approach by training the self-supervised Noise2Inverse denoising task with X-ray data replayed from real-world dynamic experiments.

show abstract

Real-time segmentation for tomographic imaging

Cited by 3 publications

References 28 publications

Real-time streaming tomographic reconstruction with on-demand data capturing and 3D zooming to regions of interest

Real-time streaming tomographic reconstruction with on-demand data capturing and 3D zooming to regions of interest

LEAN: graph-based pruning for convolutional neural networks by extracting longest chains

Just-in-time deep learning for real-time X-ray computed tomography

Contact Info

Product

Resources

About