Article
Computer Science, Information Systems
Simone Bonechi, Paolo Andreini, Alessandro Mecocci, Nicola Giannelli, Franco Scarselli, Eugenio Neri, Monica Bianchini, Giovanna Maria Dimitri
Summary: The automatic segmentation of the aorta using 2D convolutional neural networks and 3D CT scans as input is presented in this paper. A semi-automated approach was used to obtain 3D annotations for a set of CT images, and two different network architectures were compared for segmentation on three CT views. The results show promising accuracy and efficiency of the neural networks in providing aortic segmentation.
Article
Engineering, Chemical
A. Roslin, M. Lebedev, T. R. Mitchell, I. A. Onederra, C. R. Leonardi
Summary: X-ray micro-computed tomography (micro-CT) is a standard method for three-dimensional analysis of rock samples. Recent success in implementing deep learning algorithms demonstrated the potential to significantly enhance the resolution of micro-CT images. In this research, a super-resolution technique employing the U-Net 3D CNN architecture is applied to improve the resolution of granodiorite rock sample images obtained by two different 3D scanning machines.
MINERALS ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Sara Atito Ali Ahmed, Mehmet Can Yavuz, Mehmet Umut Sen, Faith Gulsen, Onur Tutar, Bora Korkmazer, Cesur Samanci, Sabri Sirolu, Rauf Hamid, Ali Ergun Eryurekli, Toghrul Mammadov, Berrin Yanikoglu
Summary: Researchers have proposed a deep learning ensemble model, IST-CovNet, for detecting COVID-19 using computed tomography (CT) or radiography images. The model achieved good performance in terms of accuracy and AUC score on different datasets. Moreover, the model has been deployed at Istanbul University's medical school for automatic screening of CT scans.
Article
Engineering, Chemical
A. Roslin, M. Marsh, N. Piche, B. Provencher, T. R. Mitchell, I. A. Onederra, C. R. Leonardi
Summary: This study proposes a super-resolution processing technique based on the three-dimensional U-Net convolutional neural network to enhance the resolution of micro-CT images of micro-porous rock samples without reducing the sample size. Super-resolution processing can significantly improve the quality of low-resolution micro-CT images. This technique is expected to help reveal features absent in small samples and save the cost and time required for high-resolution scanning.
MINERALS ENGINEERING
(2022)
Article
Computer Science, Interdisciplinary Applications
Bochao Zhao, Nishank Saxena, Ronny Hofmann, Chaitanya Pradhan, Amie Hows
Summary: The current hardware configuration of micro-CT detectors limits the resolution of rock pores that can be achieved. Super resolution techniques and image quality evaluation based on pore throat resolution are proposed to overcome this limitation. Additionally, the use of registered micro-CT images and the ratio of pore throat size to voxel size as grouping criteria in the training dataset are suggested to improve image sharpness and contrast while refining voxel size beyond current technology capabilities.
COMPUTERS & GEOSCIENCES
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Redona Brahimetaj, Jan Cornelis, Chantal Van Ongeval, Johan De Mey, Bart Jansen
Summary: The study aims to explore the impact of high-resolution 3D micro-CT breast microcalcification (MCs) images in diagnosing breast cancer. The results show that the highest classification results are obtained at the highest resolution (8 μm), which could potentially improve the diagnosis of breast cancer.
Article
Oncology
Redona Brahimetaj, Inneke Willekens, Annelien Massart, Ramses Forsyth, Jan Cornelis, Johan De Mey, Bart Jansen
Summary: In this study, a high resolution 3D scanner was used to scan breast tissue microcalcifications, and the association between their shape and texture features with malignancy was analyzed. The study also evaluated the potential of microcalcifications in diagnosing benign and malignant patients. Results showed that texture features of microcalcifications extracted in transform domains had higher discriminating power in classifying benign/malignant individual microcalcifications and samples compared to pure shape features.
Article
Computer Science, Artificial Intelligence
Ying Da Wang, Mehdi Shabaninejad, Ryan T. Armstrong, Peyman Mostaghimi
Summary: The study tested the performance of 4 CNN architectures for segmentation of 3D micro-CT images of rock samples, and introduced a new hybrid U-Net and ResNet network architecture to improve accuracy.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Information Systems
Jiancheng Yang, Xiaoyang Huang, Yi He, Jingwei Xu, Canqian Yang, Guozheng Xu, Bingbing Ni
Summary: This study introduces a new ACS convolution method that bridges the gap between 2D and 3D convolutions, utilizing pretrained weights on 2D datasets for 3D representation learning. Extensive experiments demonstrate the consistent superiority of pretrained ACS CNNs over 2D/3D CNN counterparts.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Humanities, Multidisciplinary
Chandra L. Reedy, Cara L. Reedy
Summary: The study of pores in historic bricks is important for understanding brick materials, evaluating deterioration, predicting future weathering, studying the effectiveness of protective measures, and analyzing the effects of cleaning treatments. High-resolution micro-CT combined with 3D image analysis is a promising approach for studying pores in bricks. The technique involves acquiring multiple X-ray projection images and reconstructing a 3D volume using computer algorithms, allowing for qualitative and quantitative analysis of the pore systems in bricks.
Article
Engineering, Chemical
Kostas Giannis, Christoph Thon, Guoqing Yang, Arno Kwade, Carsten Schilde
Summary: This study presents a 3D convolutional neural network (3D-CNN) methodology for generating realistic 3D models of particles. The method trains on 2D projections of particle images to predict their 3D shapes, and evaluates the accuracy of the predictions using Fourier shape descriptors (FSDs). This methodology has wide applications in particle shape analysis.
Article
Computer Science, Artificial Intelligence
Rongjun Ge, Yuting He, Cong Xia, Chenchu Xu, Weiya Sun, Guanyu Yang, Junru Li, Zhihua Wang, Hailing Yu, Daoqiang Zhang, Yang Chen, Limin Luo, Shuo Li, Yinsu Zhu
Summary: 2D X-CTRSNet is a powerful tool for reconstructing and segmenting 3D cervical vertebra CT from 2D X-ray images, providing accurate anatomical information and stereo structures. Its application has great potential in clinical imaging and diagnosis.
KNOWLEDGE-BASED SYSTEMS
(2022)
Article
Engineering, Biomedical
Huaying Liu, Guanzhong Gong, Wei Zou, Nan Hu, Jiajun Wang
Summary: This study aims to develop a framework for accurate local registration of organs in 3D CT images while preserving the topology of transformation. The Faster R-CNN method is used to detect local areas containing organs, and a weakly supervised deep neural network is used for registration. A novel 3D channel coordinate attention (CA) module is introduced to reduce the loss of position information. Experimental results show that the proposed method significantly improves the registration accuracy of organs and improves the topological preservation of transformations.
PHYSICS IN MEDICINE AND BIOLOGY
(2023)
Article
Computer Science, Interdisciplinary Applications
Stephan Gaerttner, Faruk O. Alpak, Andreas Meier, Nadja Ray, Florian Frank
Summary: This paper proposes a novel methodology for permeability prediction from micro-CT scans of geological rock samples. By using direct numerical simulation (DNS) to solve the stationary Stokes equation, the convergence issues of lattice Boltzmann methods (LBM) on complex pore geometries are circumvented, resulting in improved generality and accuracy of the training data set. The physics-informed CNN (PhyCNN) is trained using DNS-computed permeabilities and additional information about confined structures, achieving high prediction accuracy.
COMPUTATIONAL GEOSCIENCES
(2023)
Article
Robotics
Mehmet Efe Tiryaki, Sinan Ozgun Demir, Metin Sitti
Summary: In this study, a deep learning-based 3D magnetic microrobot tracking method using 2D MR images during microrobot motion is proposed. The experimental results demonstrate high accuracy in localization and visibility classification, and a 60% improvement in tracking accuracy compared to previous studies.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Energy & Fuels
Fatimah Alzubaidi, Patrick Makuluni, Stuart R. Clark, Jan Erik Lie, Peyman Mostaghimi, Ryan T. Armstrong
Summary: This paper introduces a machine learning-based approach for automatic fracture recognition from unwrapped drill-core images. The method applies a state-of-the-art convolutional neural network for object identification and segmentation, and investigates the feasibility of using synthetic fracture images for training the learning model.
JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING
(2022)
Article
Engineering, Chemical
Syeda Rubaiya Muin, Patrick T. Spicer, Kunning Tang, Yufu Niu, Maryam Hosseini, Peyman Mostaghimi, Ryan T. Armstrong
Summary: The article introduces a method of using micro-CT technology to conduct three-dimensional microstructural analysis of foam, and improves the image collection and analysis capabilities through deep learning. The study evaluates the stability mechanism of microfibrillar cellulose in foam, explains the effect of fiber entrapment on foam structure stability, and provides detailed analysis of the data and evaluation of deep learning methods.
CHEMICAL ENGINEERING SCIENCE
(2022)
Article
Chemistry, Physical
Chenhao Sun, James McClure, Steffen Berg, Peyman Mostaghimi, Ryan T. Armstrong
Summary: This article proposes a universal description of wetting on multiscale surfaces through the combination of integral geometry and thermodynamic laws. The theoretical framework is presented and applied to different limiting cases. Simulations of fluid droplets on structurally rough and chemically heterogeneous surfaces are conducted to explore the wetting behavior. The findings reveal the origin of classical wetting models within the proposed framework.
JOURNAL OF COLLOID AND INTERFACE SCIENCE
(2022)
Article
Computer Science, Interdisciplinary Applications
Kunning Tang, Quentin Meyer, Robin White, Ryan T. Armstrong, Peyman Mostaghimi, Ying Da Wang, Shiyang Liu, Chuan Zhao, Klaus Regenauer-Lieb, Patrick Kin Man Tung
Summary: This study demonstrates the benefits of using convolutional neural networks (CNNs) in accurately classifying different materials of proton exchange membrane fuel cells using X-ray micro-computed tomography. The study shows that a novel UResNet CNN can effectively segment the complete volume of the fuel cells with high accuracy. The CNN outperforms the manual segmentation, especially in separating carbon fibres and binder phase in the gas diffusion layer. Additionally, the CNN provides realistic permeability calculation results for the binder void space.
COMPUTERS & CHEMICAL ENGINEERING
(2022)
Article
Engineering, Chemical
Kunning Tang, Ying Da Wang, Peyman Mostaghimi, Mark Knackstedt, Chad Hargrave, Ryan T. Armstrong
Summary: Mineral Liberation Analysis (MLA) is an automated system for identifying minerals in mineral samples. This study extends MLA to three dimensions using 3D X-ray microcomputed tomography and convolutional neural networks. Comparison between 2D and 3D analysis shows significant differences in the results.
MINERALS ENGINEERING
(2022)
Article
Engineering, Chemical
Naif J. Alqahtani, Yufu Niu, Ying Da Wang, Traiwit Chung, Zakhar Lanetc, Aleksandr Zhuravljov, Ryan T. Armstrong, Peyman Mostaghimi
Summary: Reliable quantitative analysis of digital rock images, especially in heterogeneous rocks like carbonates with complex pore size distributions, requires precise segmentation and identification. In recent years, deep learning algorithms have provided efficient and automated solutions for digital rock super-resolution and segmentation. This study presents a framework that uses convolutional neural networks (CNNs) to achieve super-resolved segmentations of carbonate rock images in order to identify sub-resolution porosity. Comparison of voxel-wise segmentation accuracy metrics, topological features, and effective properties confirms the accuracy of the trained model and highlights the value of integrating deep learning frameworks in digital rock analysis.
TRANSPORT IN POROUS MEDIA
(2022)
Article
Physics, Applied
Kunning Tang, Ying Da Wang, James McClure, Cheng Chen, Peyman Mostaghimi, Ryan T. Armstrong
Summary: Mitigating greenhouse gas emissions and finding solutions for future energy is essential. Understanding geochemical reactions and flow behavior at the interface is important for successful underground storage. This study introduces an image-processing workflow using synchrotron-based μCT and CNNs to extract quantitative data and analyze porous materials for energy applications. The results show the need for comprehensive assessment beyond pixel-wise accuracy.
PHYSICAL REVIEW APPLIED
(2022)
Article
Energy & Fuels
Zakhar Lanetc, Aleksandr Zhuravljov, Yu Jing, Ryan T. Armstrong, Peyman Mostaghimi
Summary: Modelling multiphase flow in fractured media is a challenging problem of great importance for oil production. The use of dynamic pore network models allows for capturing the transient nature of multiphase flow, overcoming the limitations of numerical oscillations and interface tracking. The proposed hybrid numerical method offers a promising approach for investigating transient immiscible multiphase flow phenomena in fractures.
Article
Green & Sustainable Science & Technology
Elizabeth J. H. Kimbrel, Dorthe Wildenschild, Anna L. Herring, Ryan T. Armstrong
Summary: This article investigates the efficiency of CO2 trapping in geological storage. Through experiments on the imbibition and drainage processes of proxy fluids, it is found that the amount of trapped CO2 is dependent on the presence of the nonwetting phase and the initial injection method of supercritical CO2, factors that are not considered in current trapping models.
INTERNATIONAL JOURNAL OF GREENHOUSE GAS CONTROL
(2022)
Article
Mechanics
James E. McClure, Ming Fan, Steffen Berg, Ryan T. T. Armstrong, Carl Fredrik Berg, Zhe Li, Thomas Ramstad
Summary: Relative permeability is derived from conservation of energy and used to model fluid flow through porous materials. The study finds dynamic connectivity and explores the distribution of energy fluctuations during steady-state flow. It demonstrates the effectiveness of the conventional relative permeability relationship in simulating energy dissipation in systems with complex pore-scale dynamics.
Article
Engineering, Chemical
Yufu Niu, Samuel J. Jackson, Naif Alqahtani, Peyman Mostaghimi, Ryan T. Armstrong
Summary: This study compares two state-of-the-art super-resolution deep learning techniques, demonstrating that the unpaired GAN approach can accurately reconstruct super-resolution images as precise as the paired CNN method, with comparable training times and dataset requirements. This opens up new possibilities for micro-CT image enhancement using unpaired deep learning methods, eliminating the need for image registration.
TRANSPORT IN POROUS MEDIA
(2022)
Article
Environmental Sciences
Mohammad Ebadi, Ryan T. Armstrong, Peyman Mostaghimi, Ying Da Wang, Naif Alqahtani, Tammy Amirian, Lesley Anne James, Arvind Parmar, David Zahra, Hasar Hamze, Dmitry Koroteev
Summary: This paper proposes a fully automated workflow based on soft computing to characterize the heterogeneous flow properties of cores and build predictive continuum-scale models. The workflow classifies rock types using image features and morphological properties, and evaluates petrophysical properties through pore-scale simulations. Experimental results demonstrate the high-fidelity characterization provided by this workflow.
WATER RESOURCES RESEARCH
(2022)
Article
Chemistry, Physical
Quentin Meyer, Shiyang Liu, Karin Ching, Ying Da Wang, Chuan Zhao
Summary: This study introduces a new approach to monitoring the evolutions of the triple-phase boundary in proton exchange membrane fuel cells. By simultaneously capturing the ohmic resistance and double-layer capacitance using high-frequency zero-phase impedance spectroscopy, the impact of fuel cell operations on the triple-phase boundary can be accurately analyzed. The results show that the triple-phase boundary area reduces throughout the operations, indicating changes in proton transport, electron transport, and oxygen diffusion.
JOURNAL OF POWER SOURCES
(2023)
Article
Engineering, Chemical
Kunning Tang, Ying Da Wang, Yufu Niu, Tom A. Honeyands, Damien O'Dea, Peyman Mostaghimi, Ryan T. Armstrong, Mark Knackstedt
Summary: The iron ore sintering process needs to be optimized to reduce energy consumption and carbon emissions, while producing high-quality sinter for low carbon blast furnace operations. 3D X-ray micro-computed tomography can provide particle-level information of iron ore mixtures, but algorithms are needed to identify and classify particles and establish the relationship between ore sources and sinter quality.
Article
Multidisciplinary Sciences
Ying Da Wang, Quentin Meyer, Kunning Tang, James E. McClure, Robin T. White, Stephen T. Kelly, Matthew M. Crawford, Francesco Iacoviello, Dan J. L. Brett, Paul R. Shearing, Peyman Mostaghimi, Chuan Zhao, Ryan T. Armstrong
Summary: The authors utilize X-ray micro-computed tomography, deep learned super-resolution, multi-label segmentation, and direct multiphase simulation to simulate fuel cells and guide their design, addressing the challenge of accurate liquid water modelling.
NATURE COMMUNICATIONS
(2023)