High-fidelity geometry generation from CT data using convolutional neural networks

Tekawade, Aniket; Sforzo, Brandon; Matusik, Katarzyna E.; Kastengren, Alan L.; Powell, Christopher F.

doi:10.1117/12.2540442

Cited by 12 publications

(14 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Ulyanov et al [20] investigated features of the U-net architecture and noted that the deepest parts of the analysis (converging) path encoded the most highlevel context, thereby enabling tasks such as in-painting and denoising. We have shown in previous work [5] that if Unet is too deep (more max-pooling layers), it overfits on the shapes observed in the training data, leading to models that do not generalize accurately. Here we exploit this weakness to our advantage and investigate if shape priors (i.e., porosity) can be systematically learnt in the deepest part.…”

Section: Autoencoder Architecture and Trainingmentioning

confidence: 99%

“…A typical approach to extracting porosity metrics starts with semantic segmentation (popularly using U-net or similar encoder-decoder architectures [1,2,3,4,5] where voxels, classified as either void or material (substrate) based on the local intensity mapping, are assigned values of 0 and 1 respectively. Pores that are spatially disconnected from others are then uniquely labelled by applying a connected component filter (instance segmentation) and the aforementioned statistical porosity metrics ("features") are calculated.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

3d Autoencoders For Feature Extraction In X-Ray Tomography

Tekawade

Liu

Kenesei

et al. 2021

2021 IEEE International Conference on Image Processing (ICIP)

Self Cite

View full text Add to dashboard Cite

Real-time steering of time-resolved or in-situ X-ray tomography requires capturing changes in morphological descriptors in a sample (e.g., porosity, particle size, and crack width) during continuous data acquisition. Segmentation of 2D or 3D images followed by quantitative measurement is the conventional method for tracking changes in these descriptors with respect to a previous time-step or a 3D search in a volume. However, image segmentation is expensive. As a faster and unsupervised alternative, we propose a feature-extraction approach using a convolutional autoencoder, where the latent space of the encoder responds to relative changes in morphology without prior knowledge of the morphological descriptors. To test this approach in parametric experiments, we used a digital twin for micro-CT to generate realistic datasets of porous materials. We show that for some configurations of the autoencoder, its n-dimensional latent space encodes a sample's local porosity metrics while disregarding contrast information (relative proportion of absorption and phase contrast) determined by the imaging modality and not the sample. Through dimensionality reduction, the vector's response is visualized in 2D space to find clusters of data with similar porosity metrics. Our approach can extract features from grayscale tomographic data more than 4x faster than a segmentation + pore analysis workflow.

show abstract

Section: Autoencoder Architecture and Trainingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

3d Autoencoders For Feature Extraction In X-Ray Tomography

Tekawade

Liu

Kenesei

et al. 2021

2021 IEEE International Conference on Image Processing (ICIP)

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, instead of featuring five holes, the A-M3 injector has three side-oriented holes that are nominally oriented at an angle of 73° with respect to the needle axis and are characterized by a sharp inlet radius of curvature to promote cavitation. The eroded injector surfaces were generated from the x-ray image analysis workflow described in Tekawade et al [6]. A comparison of the baseline and eroded injectors is shown in Figure 1(a) and (b), respectively, where details from the machining process can be seen in the sac and erosion patterns are evident among the three orifices.…”

Section: Computational Model Set-upmentioning

confidence: 99%

Linking cavitation erosion in a multi-hole injector with spray and combustion development

Magnotti

Kundu

Nunno

et al. 2021

ICLASS

View full text Add to dashboard Cite

It is well known that cavitation erosion in fuel injectors can prevent reliable engine performance after only several thousand hours of operation. However, current simulation tools lack the ability to link flow predictions within the fuel injector to both the efficacy of combustion strategies and lifetime of the injector. Multiphase flow simulation predictions were studied and compared between an informed baseline injector geometry and an x-ray scanned eroded injector geometry. Overall, erosion was found to decrease the fuel mass delivery and injection velocities. A two-stage static coupling approach was employed to link the predicted injection conditions from non-eroded and eroded injectors with the external spray simulations under reacting conditions. Combustion modeling in this coupled approach was carried out using the Unsteady Flamelet Progress Variable approach with a detailed chemical mechanism for ndodecane, comprising of 2,755 species and 11,173 reactions. Erosion in the injectors led to lower rates of spray penetration in comparison to the baseline configurations. Analysis of the reacting spray simulations revealed an insensitivity of ignition to erosion, yet shorter lift off lengths and higher levels of the soot precursor acetylene were predicted in the eroded injector.

show abstract

“…This poses a challenge to studies seeking to thoroughly investigate large, complex samples at high resolution, ranging from disease models (e.g. developing mice, zebrafish, Daphnia ), to micro-circuits in electronic components, to fuel spray systems (Wong et al ., 2012;Ding et al ., 2019; De Samber et al ., 2008; Lall and Wei, 2015; Tekawade et al ., 2019).…”

Section: Introductionmentioning

confidence: 99%

A Wide-Field Micro-Computed Tomography Detector: Micron Resolution at Half-centimeter Scale

Yakovlev

Vanselow²,

Ngu³

et al. 2021

Preprint

View full text Add to dashboard Cite

Ideal 3-dimensional imaging of many complex samples, such as biological tissues made up of micro-scale structures extending over millimeter- to centimeter-scale tissue samples and organisms, requires both a wide field-of-view and high resolution. With existing optics and detectors used for micro-CT imaging, sub-micron pixel resolution can only be achieved for fields-of-view of <2 mm. This manuscript presents a unique detector system with a 6-mm field-of-view image circle and 0.5 μm pixel size that can be used in both synchrotron facilities and tabletop micro-CT units. A resolution-test pattern with linear microstructures and whole adult Daphnia magna were imaged on Beamline 8.3.2 of the Advanced Light Source. Volumes of 10,000 × 10,000 × 7,096 isotropic 0.5 μm voxels were reconstructed over a 5.0 × 3.5 mm field-of-view. Measurements in the projection domain confirmed a 1.182 μm measured spatial resolution that is largely Nyquist-limited. This unprecedented combination of field-of-view and resolution dramatically reduces the need for sectional scans and computational stitching for large samples, ultimately offering the means to elucidate change in tissue and cellular morphology in the context of larger whole, intact model organisms and specimens. This development is also anticipated to benefit micro-CT imaging in materials science, microelectronics, agricultural science, and biomedical engineering.

show abstract

High-fidelity geometry generation from CT data using convolutional neural networks

Cited by 12 publications

References 10 publications

3d Autoencoders For Feature Extraction In X-Ray Tomography

3d Autoencoders For Feature Extraction In X-Ray Tomography

Linking cavitation erosion in a multi-hole injector with spray and combustion development

A Wide-Field Micro-Computed Tomography Detector: Micron Resolution at Half-centimeter Scale

Contact Info

Product

Resources

About