Article
Computer Science, Artificial Intelligence
Junwei Han, Yang Yang, Dingwen Zhang, Dong Huang, Dong Xu, Fernando De La Torre
Summary: This paper focuses on learning category-specific 3D shape models under weak supervision, where only object categories and keypoints are manually annotated on training 2D images, to improve object segmentation performance. Confidence weighting schemes are developed to reduce confusion caused by noisy data and increase chances for recovering reliable 3D object shapes.
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
(2021)
Article
Computer Science, Artificial Intelligence
Yang Yang, Junwei Han, Dingwen Zhang, Qi Tian
Summary: This study proposes a method for 3D shape reconstruction using rich intermediate representations through a newly designed network architecture. By utilizing a two-stream network and a shape transformation network, detailed features of the entire 3D object shapes are successfully reconstructed.
PATTERN RECOGNITION
(2022)
Article
Geochemistry & Geophysics
Zhuan Dai, Jianhua Hu, Shaowei Ma, Yaguang Qin, Xiao Xu
Summary: In this paper, an improved grain-based model is proposed to construct random mineral grains with real 3D shape in numerical specimens. By using CT scan and image processing techniques, the complex 3D shape of minerals is reconstructed, and then, random grains with real shape are generated. Compression tests on the numerical specimens show that their mechanical properties are similar to those of real specimens. Therefore, the proposed method is feasible and reasonable.
Article
Biology
Jorg Sander, Bob D. de Vos, Steffen Bruns, Nils Planken, Max A. Viergever, Tim Leiner, Ivana Isgum
Summary: Since the advent of computer-aided diagnosis in medical imaging, voxel-based segmentation has become the primary method for analyzing left ventricle (LV) function and morphology in cardiac MR images. However, traditional 2D MR imaging suffers from limitations in obtaining accurate 3D cardiac anatomy due to anisotropy and respiratory motion. To address this, the proposed method utilizes a training auto-decoder to learn a continuous implicit function representing 3D LV shapes. The results demonstrate that this approach can reconstruct high-resolution LV shapes and correct for motion-induced artifacts, providing more accurate assessment of structural and functional abnormalities.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Computer Science, Software Engineering
Mahdi Abbaspour Tehrani, M. Gopi, Aditi Majumder
Summary: This article presents an automated multi-projector registration system for multiple uncalibrated projectors and cameras on arbitrary surfaces. The method estimates parameters, registers projectors geometrically without fiducials, and allows for quick recalibration with projector movements. This system enables easy deployment of spatially augmented reality environments of various sizes, shapes, and configurations.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2021)
Article
Multidisciplinary Sciences
Jonathan Schwartz, Chris Harris, Jacob Pietryga, Huihuo Zheng, Prashant Kumar, Anastasiia Visheratina, Nicholas A. Kotov, Brianna Major, Patrick Avery, Peter Ercius, Utkarsh Ayachit, Berk Geveci, David A. Muller, Alessandro Genova, Yi Jiang, Marcus Hanwell, Robert Hovden
Summary: The authors demonstrate real-time tomography with dynamic 3D tomographic visualization integrated in tomviz, enabling rapid interpretation of specimen structure and obtaining high-quality tomograms in a short period of time.
NATURE COMMUNICATIONS
(2022)
Article
Polymer Science
Ioana Caloian, Jocelyn Trapp, Melissa W. Williams, Ryan A. Kim, Mahmoud E. Moustafa, Eva Hawa Stwodah, Christina Tang
Summary: In this work, the authors demonstrate the simultaneous patterning of fibers and fabrication of functional 2D and 3D shapes using a one-step electrospinning process. The use of mesh templates allows for preferential attraction of electrospun fibers to metal protrusions, resulting in a patterned mat that mimics a woven mesh. This approach offers potential applications in personalized wound care and surgical meshes.
Article
Biochemical Research Methods
Qiufu Li, Linlin Shen
Summary: The paper introduces a 3D wavelet and deep learning-based method for neuron segmentation, utilizing 3D WaveUNet to process neuronal cubes and improve performance in noisy neuronal images. The integrated 3D wavelets efficiently assist in 3D neuron segmentation and reconstruction.
Article
Multidisciplinary Sciences
Eun Jung Lee, Min Gu Kim, Mi Sun Chung, Seon-Ok Kim, Jun Soo Byun, Younghee Yim
Summary: This study aimed to evaluate the agreement between conventional pre-contrast 3D T1 MPRAGE and wave-CAIPI MPRAGE in the diagnosis of intracranial lesions. The results showed that both methods demonstrated good agreement in the diagnosis of intracranial lesions. Wave-CAIPI MPRAGE achieved a shorter scan time by approximately 50% but had poorer image quality compared to conventional MPRAGE.
SCIENTIFIC REPORTS
(2022)
Article
Materials Science, Multidisciplinary
Ilse B. Nava-Medina, Karli A. Gold, Savannah M. Cooper, Kyle Robinson, Abhishek Jain, Zhengdong Cheng, Akhilesh K. Gaharwar
Summary: The combination of self-oscillating chemical reactions with 3D printing techniques allows the construction of 3D geometries with different shapes, sizes, and angles to investigate spatial pattern formation. The study found that size variations in geometries can affect the oscillation frequencies of patterns, playing an important role in spatial alterations.
ADVANCED MATERIALS TECHNOLOGIES
(2021)
Article
Multidisciplinary Sciences
Anika Kueken, Damoun Langary, Zoran Nikoloski
Summary: Understanding the complexity of metabolic networks can be characterized by identifying multi-reaction dependencies. Concordant complexes can capture these dependencies and reduce the apparent complexity of metabolic networks. The metabolic network of Escherichia coli is more tightly coordinated than expected by chance.
Article
Computer Science, Interdisciplinary Applications
Sergey Petropavlovsky, Semyon Tsynkov, Eli Turkel
Summary: This study presents a high accuracy method for computing the scattering of unsteady acoustic waves around complex three-dimensional bodies. The geometry of the scattering body is defined using CAD and its surface is represented as non-overlapping patches parameterized independently using high order splines. The method, which reduces the wave equation to a system of Calderon's boundary operator equations, demonstrates grid-independence and sub-linear complexity. It efficiently handles complex non-conforming geometries without compromising accuracy or stability and can handle long simulation times. The exact treatment of artificial outer boundaries and the ability to efficiently solve similar problems are additional advantages.
JOURNAL OF COMPUTATIONAL PHYSICS
(2022)
Article
Gastroenterology & Hepatology
Kazuhiro Togasaki, Shinya Sugimoto, Yuki Ohta, Kosaku Nanki, Mami Matano, Sirirat Takahashi, Masayuki Fujii, Takanori Kanai, Toshiro Sato
Summary: In diffuse gastric cancer, signet-ring cell carcinoma (SRCC) and poorly cohesive carcinoma not otherwise specified (PCC-NOS) were found to be clonally identical, with their morphology regulated by external Wnt and R-spondin expression. The study results decoded how genetic mutations and the tumor microenvironment shape pathohistologic and biologic phenotypes in human diffuse gastric cancers.
Article
Chemistry, Analytical
Sukwoo Jung, Youn-Sung Lee, Yunju Lee, KyungTaek Lee
Summary: This article introduces a method of using stereo cameras and ToF sensors to improve the accuracy of 3D information in depth sensing. By employing multi-camera calibration, depth map fusion, hole-filling, and surface reconstruction, the transformation from sensor data to point cloud and mesh data is achieved.
Article
Engineering, Mechanical
Tianbao Liang, Mu He, Hao-Wen Dong, Liang Xia, Xiaodong Huang
Summary: This research proposes a novel design strategy for ultrathin and highly efficient waterborne reflective pentamode metasurfaces to achieve uniform diffuse reflections in underwater scenes. A theoretical model is established to ease the demand on impedance matching and construct an ideal diffusion field. The spatially variant equivalent impedances of the metasurface are identified, and their corresponding pentamode material configurations are inversely designed with band structure analyses. Numerical results show high performance at the targeted frequency, and further verifications reveal applicability to a broader frequency range, paving the way for deep subwavelength scale acoustic wave manipulations with ultrathin waterborne metasurfaces.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)