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
Environmental Sciences
Mingzhe Feng, Xin Sun, Junyu Dong, Haoran Zhao
Summary: This paper proposes a network structure that uses dynamic receptive field and Gaussian pyramid pooling to address the issue of scale variation in remote sensing image segmentation. The network achieves better performance than other methods on large remote sensing image datasets.
Article
Computer Science, Interdisciplinary Applications
Hans Pinckaers, Wouter Bulten, Jeroen van der Laak, Geert Litjens
Summary: Prostate cancer is the most prevalent cancer among men in Western countries, and pathologists' evaluation is the gold standard for diagnosis. State-of-the-art convolutional neural networks are often patch-based and require detailed pixel-level annotations for effective training.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Computer Science, Interdisciplinary Applications
Javier Civit-Masot, Alejandro Banuls-Beaterio, Manuel Dominguez-Morales, Manuel Rivas-Perez, Luis Munoz-Saavedra, Jose M. Rodriguez Corral
Summary: This study designs, implements, and evaluates a diagnostic aid system for non-small cell lung cancer detection using Deep Learning techniques. The results show high accuracy and area under the ROC curve, indicating improved classification results and reduced time for pathologists and diagnostic turnaround time.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Article
Computer Science, Software Engineering
Yi Xiao, Jin Wu, Jie Zhang, Peiyao Zhou, Yan Zheng, Chi-Sing Leung, Ladislav Kavan
Summary: This article introduces a deep learning-based method for generating colored images, which allows users to control the results through global and local inputs. The authors propose a two-stage deep colorization method and design a loss function to differentiate the influences of different inputs. They also propose the application of color theme recommendation and image compression scheme.
IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS
(2022)
Article
Chemistry, Analytical
Darian Tomasevic, Peter Peer, Franc Solina, Ales Jaklic, Vitomir Struc
Summary: This paper investigates the task of reconstructing 3D scenes based on visual data using superquadrics as volumetric primitives, with an extension to intensity and color images. The research shows that a dedicated convolutional neural network model can accurately reconstruct superquadrics, not only for simple object images, but also for more complex images with textures, outperforming the current state-of-the-art method.
Article
Automation & Control Systems
Wenqiang Li, Yuk Ming Tang, Ziyang Wang, Kai Ming Yu, Suet To
Summary: Automatic vertebrae segmentation using CT plays a crucial role in automated spine analysis, and recent advancements in deep learning have led to precise performance through deep convolutional neural networks. While DCNN-based semantic segmentation algorithms have advantages, they face limitations that are addressed by the proposed novel algorithm, which includes encoder-decoder framework, Layer Normalization, Atrous Residual Path, and a 3D Attention Module to improve segmentation accuracy. Experimental results show competitive performance compared to existing methods for automatic vertebrae semantic segmentation.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Environmental Sciences
Guanzhou Chen, Xiaoliang Tan, Beibei Guo, Kun Zhu, Puyun Liao, Tong Wang, Qing Wang, Xiaodong Zhang
Summary: Semantic segmentation is a fundamental task in remote sensing image analysis, and our proposed SDFCNv2 framework shows better performance on remote sensing images compared to the SDFCNv1 framework, increasing the mIoU metric by up to 5.22% while using only about half of the parameters.
Article
Computer Science, Theory & Methods
Hamidreza Bolhasani, Somayyeh Jafarali Jassbi, Arash Sharifi
Summary: The powerful capability of deep learning in image classification problems has gained popularity and application in various fields, including medical sciences. In real-time medical applications, accuracy is not the only concern, but computation runtime and power consumption are equally important performance indicators.
JOURNAL OF BIG DATA
(2023)
Article
Engineering, Electrical & Electronic
Philipp Schuegraf, Stefano Zorzi, Friedrich Fraundorfer, Ksenia Bittner
Summary: Urban areas are made up of complex building structures, which can be seen in high-resolution remote sensing images, including both the buildings and the lines that separate them. This study proposes a two-stage approach to address the limitations of previous building segmentation methods. The first stage uses an FCN model to segment the buildings and the separation lines, while the second stage uses a learning-free method to generate building instances.
IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING
(2023)
Article
Public, Environmental & Occupational Health
Abdolreza Marefat, Mahdieh Marefat, Javad Hassannataj Joloudari, Mohammad Ali Nematollahi, Reza Lashgari
Summary: COVID-19 is a novel virus that rapidly spreads and affects individuals' lives in various ways. Detecting the virus is crucial, and medical imaging such as CT and X-ray images are commonly used. However, the current procedures and high caseloads present challenges for medical practitioners. In this study, we propose a transformer-based method using Compact Convolutional Transformers (CCT) for automatically detecting COVID-19 from X-ray images. Our experiments demonstrate the effectiveness of the method with an accuracy of 99.22%, outperforming previous works.
FRONTIERS IN PUBLIC HEALTH
(2023)
Article
Computer Science, Artificial Intelligence
Shohreh Sheiati, Sanaz Behboodi, Navid Ranjbar
Summary: This study evaluates the potential of SegNet, a deep autoencoder convolutional network, for automatic segmentation of fly ash-based geopolymer images. The results show that SegNet achieves comparable accuracy to human performance even with a few training images, and it is able to adapt itself to different magnification levels. Compared to the Gaussian method, SegNet outperforms in uncontrolled imaging conditions and demonstrates self-learning capability in poorly annotated areas.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Shohreh Sheiati, Sanaz Behboodi, Navid Ranjbar
Summary: This study evaluates the potential of deep autoencoder convolutional network SegNet in automatic segmentation of backscattered electron images, demonstrating its ability to achieve comparable accuracy to human performance, as well as magnification independent training and self-learning capability.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Yide Di, Xiaoke Zhu, Xin Jin, Qiwei Dou, Wei Zhou, Qing Duan
Summary: This paper introduces a deep convolutional network framework called Color-UNet++ for colorization problems, addressing issues such as gradient dispersion and uneven overlap. By adjusting the model structure, color space, and objective function, the method proves to be superior through extensive experimental results on LFW and LSUN datasets.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Oncology
Lei Jin, Feng Shi, Qiuping Chun, Hong Chen, Yixin Ma, Shuai Wu, N. U. Farrukh Hameed, Chunming Mei, Junfeng Lu, Jun Zhang, Abudumijiti Aibaidula, Dinggang Shen, Jinsong Wu
Summary: This study utilized deep learning for glioma classification and developed a neuropathological diagnostic platform that can accurately identify glioma histological subtypes using CNNs, with accuracy rates up to 86.5%-87.5%. Furthermore, by incorporating molecular markers, histopathological classifications could be further extended to integrated neuropathological diagnosis.
Article
Computer Science, Artificial Intelligence
Chao Ma, Minjie Wan, Yunkai Xu, Kan Ren, Weixian Qian, Qian Chen, Guohua Gu
Summary: This paper proposes an infrared target tracker based on the proximal robust principal component analysis method, solves the convex optimization problem using the Alternating Direction Method of Multipliers, and locates the target using the particle filter framework.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Kan Ren, Yuan Gao, Minjie Wan, Guohua Gu, Qian Chen
Summary: Infrared small target detection has always been a challenging problem due to the limited pixels and features of small infrared targets. Current optimization methods primarily focus on multi-scale feature fusion or super-resolution enhancement. However, when applying super-resolution networks to infrared target detection, two significant issues arise: excessive computational power consumption, resulting in low detection rates, and the disparity between low-resolution training images and the actual distribution of tiny targets, leading to poor detection accuracy. This paper proposes a new detection network, RSRGAN, which consists of a computationally efficient backbone network (RCN) for extracting potential regions and a GAN-based generator that includes modules for distribution transformation and super-resolution enhancement. The discriminator assists in generating better super-resolution images by distinguishing between generated and actual images. Additionally, the authors create an infrared UAV small target dataset, which includes birds, leaves, and other disturbances, to evaluate the algorithm's performance. Experimental results demonstrate that the proposed method achieves better detection of small IR targets and outperforms existing approaches.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Artificial Intelligence
Wuxin Li, Qian Chen, Guohua Gu, Xiubao Sui
Summary: The study proposes a method for object matching between visible and infrared images using a Siamese neural network combined with a convolutional neural network for feature extraction and cross-correlation calculations for matched targets, achieving higher accuracy and precision in experiments.
APPLIED INTELLIGENCE
(2022)
Article
Geochemistry & Geophysics
Minjie Wan, Guohua Gu, Yunkai Xu, Weixian Qian, Kan Ren, Qian Chen
Summary: This paper proposes a total variation-based interframe infrared patch-image model for infrared small target detection. By converting the infrared image into a patch-image consisting of a sparse target matrix and a low-rank background matrix, and introducing temporal consistency constraint and TV regularization term, the proposed model achieves better performance in infrared small target detection.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Minjie Wan, Xiaobo Ye, Xiaojie Zhang, Yunkai Xu, Guohua Gu, Qian Chen
Summary: The precision of infrared small target tracking is improved by a new method using compressive convolution feature extraction. This method utilizes a Gaussian curvature feature map and a three-layer compressive convolution network to represent each candidate target. Experiments show that the method achieves more satisfactory performances compared with other typical visual trackers.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Biochemical Research Methods
Shun Zhou, Jiaji Li, Jiasong Sun, Ning Zhou, Qian Chen, Chao Zuo
Summary: Fourier Ptychographic diffraction tomography (FPDT) is a label-free computational microscopy technique that retrieves high-resolution 3D tomograms by synthesizing low-resolution intensity images. Accelerated Fourier Ptychographic diffraction tomography (aFPDT) combines sparse annular LED illuminations and multiplexing to significantly reduce data and achieve computational acceleration. The method has been experimentally demonstrated on biological cells and has the potential to advance applications in biomedicine.
JOURNAL OF BIOPHOTONICS
(2022)
Review
Dentistry, Oral Surgery & Medicine
Y. Xu, X. Deng, Y. Sun, X. Wang, Y. Xiao, Y. Li, Q. Chen, L. Jiang
Summary: Oral potentially malignant disorders (OPMDs) are a group of oral lesions that have a variable risk of transforming into oral cancer. Traditional oral examinations may miss early malignant lesions, but optical techniques offer noninvasive methods for detecting these changes. This article provides a comprehensive review of various optical imaging methods used in the diagnosis of OPMDs, discussing their advantages, limitations, and future challenges.
JOURNAL OF DENTAL RESEARCH
(2022)
Article
Optics
Jiajie Wang, Minjie Wan, Guohua Gu, Weixian Qian, Kan Ren, Qinyan Huang, Qian Chen
Summary: By proposing a new underwater image restoration method based on periodic integration of polarization images to replace orthogonal polarization images in traditional PDI systems, better performance in texture enhancement and noise suppression was achieved in underwater image restoration compared to existing PDI methods based on one or two pairs of orthogonal polarization images.
OPTICS AND LASERS IN ENGINEERING
(2022)
Article
Optics
Yan Hu, Zhongwei Liang, Shijie Feng, Wei Yin, Jiaming Qian, Qian Chen, Chao Zuo
Summary: The article introduces the Scheimpflug lens-based imaging model and its application, proposes a simplified imaging model based on projection, and develops calibration algorithms and rectification methods for stereo matching. The effectiveness and accuracy of the methods are verified through experiments.
OPTICS AND LASERS IN ENGINEERING
(2022)
Review
Optics
Wenwen Zhang, Daquan Yu, Yongcheng Han, Weiji He, Qian Chen, Ruiqing He
Summary: This paper introduces a computational ghost imaging technique to estimate the depth of an object by evaluating the degree of defocus in reconstructed images. It analyzes the images formed by multi-depth objects, selects the gradient domain as the image transformation space, and uses a compressed sensing algorithm based on TV norm to form the images. Additionally, it proposes a depth estimation strategy using variable resolution speckles to reduce the required depth slices.
OPTICS AND LASERS IN ENGINEERING
(2022)
Article
Optics
Xiaochen Zhao, Xiaoduo Jiang, Aojie Han, Tianyi Mao, Weiji He, Qian Chen
Summary: This paper explores how to improve the accuracy of edge reconstruction under sparse photon conditions and investigates the correlation between edge reconstruction accuracy and overall depth reconstruction accuracy.
Article
Optics
Yixuan Li, Jiaming Qian, Shijie Feng, Qian Chen, Chao Zuo
Summary: Single-shot fringe projection profilometry (FPP) is crucial for retrieving the absolute depth information of objects in high-speed dynamic scenes. This study proposes a composite fringe projection deep learning profilometry (CDLP) method that combines deep learning and physical model to achieve high-precision and unambiguous phase retrieval on a single composite fringe image. The method overcomes the problem of serious spectrum aliasing caused by multiplexing schemes and can reconstruct high-quality absolute 3D surfaces.
Article
Optics
Ning Zhou, Jiaji Li, Runnan Zhang, Zhidong Bai, Shun Zhou, Qian Chen, Chao Zuo
Summary: This study presents a 3D label-free refractive index imaging technique based on single-exposure intensity diffraction tomography (sIDT) using a color-multiplexed illumination scheme. By capturing the scattering field from different directions and separating monochromatic intensity images, the 3D refractive index distribution of the object can be reconstructed. The method demonstrates reliable performance in label-free, high-throughput, and real-time 3D volumetric biological imaging applications.
Article
Optics
Xueqi Chen, Lin Zhou, Meng Zhou, Ajun Shao, Kan Ren, Qian Chen, Guohua Gu, Minjie Wan
Summary: This paper proposes an infrared ocean image simulation algorithm based on the Pierson-Moskowitz spectrum and a bidirectional reflectance distribution function, which can provide more authentic and clear simulation images.
Article
Optics
Miao Wu, Yu Lu, Haochen Li, Tianyi Mao, Yanqiu Guan, Labao Zhang, Weiji He, Peiheng Wu, Qian Chen
Summary: Long-range lidar systems often record large but extremely sparse data cubes, making it challenging to accurately estimate depth images. This paper introduces an intensity-guided method that utilizes temporal-spatial correlation to estimate depth images. Preprocessing steps are used to reduce the data size. Experimental results show that this method outperforms other state-of-the-art methods in estimating depth images, particularly in low signal return scenarios.
OPTICS AND LASERS IN ENGINEERING
(2022)
Article
Instruments & Instrumentation
Zengrun Wen, Xiulin Fan, Kaile Wang, Weiming Wang, Song Gao, Wenjing Hao, Yuanmei Gao, Yangjian Cai, Liren Zheng
Summary: This study presents a transition from Q-switched state to continuous wave state in an erbium-doped fiber laser, accompanied by irregular mode-hopping. The results showed that the transition between these two states can be achieved by adjusting the pump power. Modulation peaks were observed on both the Q-switched pulse train and the continuous wave background, and the central wavelength fluctuated.
INFRARED PHYSICS & TECHNOLOGY
(2024)