Article
Chemistry, Multidisciplinary
Roziana Ramli, Khairunnisa Hasikin, Mohd Yamani Idna Idris, Noor Khairiah A. Karim, Ainuddin Wahid Abdul Wahab
Summary: This study introduces a feature-based retinal fundus image registration technique that combines retinal vessels and noise characteristics to increase accuracy. The registration method CURVE-SIFT outperformed existing methods significantly, achieving a higher success rate in image pairing.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Interdisciplinary Applications
Dian Qin, Jia-Jun Bu, Zhe Liu, Xin Shen, Sheng Zhou, Jing-Jun Gu, Zhi-Hua Wang, Lei Wu, Hui-Fen Dai
Summary: Recent advancements in applying convolutional neural networks to medical image segmentation have led to a more precise prediction results, but existing methods often rely on high computational complexity and storage, which is not practical. To address this issue, a new efficient architecture is proposed by distilling knowledge from well-trained networks to train a lightweight network, resulting in improved segmentation capability while maintaining runtime efficiency.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Computer Science, Artificial Intelligence
Xiaofei Bian, Haiwei Pan, Kejia Zhang, Peng Liu, Chunling Chen
Summary: This paper proposes a method for the classification of malignant melanoma dermoscopy images based on multi-modal medical features, which can reduce the classification error caused by the complexity and subjectivity of visual interpretation and assist dermatologists in analyzing the skin lesion area.
IET IMAGE PROCESSING
(2023)
Article
Computer Science, Interdisciplinary Applications
Bo Huang, Ziran Wei, Xianhua Tang, Hamido Fujita, Qingping Cai, Yongbin Gao, Tao Wu, Liang Zhou
Summary: A two-dimensional deep learning segmentation network based on multi-pinacoidal plane fusion was proposed for medical volume data, achieving satisfactory progress on different backbone networks. The approach covers more information and shows compatibility while extracting global information between different input layers.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2021)
Article
Geochemistry & Geophysics
Lizwe Wandile Mdakane, Waldo Kleynhans
Summary: The study aims to develop a monitoring system for automatically detecting oil spill events caused by vessels in African Oceans and identify critical features for distinguishing oil spills from look-alikes. Investigation of common feature selection methods and classifiers reveals consistent significance of certain features across all methods.
IEEE GEOSCIENCE AND REMOTE SENSING LETTERS
(2022)
Article
Geochemistry & Geophysics
Qiaobo Hao, Bin Sun, Shutao Li, Melba M. Crawford, Xudong Kang
Summary: A curvature filters-based multiscale feature extraction method with multiscale superpixel segmentation constraint is proposed for hyperspectral image classification. Experimental results demonstrate significant improvement in classification accuracies compared to standard methods, especially in cases of limited training samples.
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING
(2022)
Article
Engineering, Electrical & Electronic
Yoshiki Nakayama, Huimin Lu, Yujie Li, Tohru Kamiya
Summary: This paper proposed a new architecture called WideSegNeXt to address the shortcomings of the FCN in semantic segmentation, which can capture image context on various spatial scales, effectively identify small objects, and reduce the loss of position information. The proposed method achieved a mean intersection over union (MIoU) of 72.5% and a global accuracy (GA) of 92.4% on the CamVid dataset without the need for additional input datasets, outperforming previous methods.
IEEE SENSORS JOURNAL
(2021)
Article
Computer Science, Information Systems
Tongping Shen, Huanqing Xu
Summary: This paper proposes a dual-encoder image segmentation network, including HarDNet68 and Transformer branch, which can extract the local features and global feature information of the input image, allowing the segmentation network to learn more image information, thus improving the effectiveness and accuracy of medical segmentation.
Article
Computer Science, Interdisciplinary Applications
Junlin Xian, Xiang Li, Dandan Tu, Senhua Zhu, Changzheng Zhang, Xiaowu Liu, Xin Li, Xin Yang
Summary: In this work, a novel unsupervised domain adaptation (UDA) method called dual adaptation-guiding network (DAG-Net) is proposed for medical image segmentation. DAG-Net incorporates two effective structural-oriented guidance modules, Fourier-based contrastive style augmentation (FCSA) and residual space alignment (RSA), to adapt a segmentation model from a labelled source domain to an unlabeled target domain. Experimental results demonstrate the superior performance of DAG-Net compared to state-of-the-art UDA approaches for 3D medical image segmentation on unlabeled target images.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2023)
Article
Engineering, Civil
Dimitrios Loverdos, Vasilis Sarhosis
Summary: This study uses computer vision and convolutional neural networks to automatically detect bricks and cracks in masonry structures, and develops a dynamic workflow for creating geometric digital twins to address the issues of structural assessment and documentation.
ENGINEERING STRUCTURES
(2023)
Article
Construction & Building Technology
Dimitrios Loverdos, Vasilis Sarhosis, Efstathios Adamopoulos, Anastasios Drougkas
Summary: The accuracy in transferring the real structure geometry to the numerical model is vital when modeling the mechanical behavior of existing masonry structures. Advances in photogrammetry and image processing have enabled the rapid and remote digital recording of objects and features. A proposed framework based on image processing automatically extracts geometrical features from masonry structures for advanced numerical modeling.
AUTOMATION IN CONSTRUCTION
(2021)
Article
Health Care Sciences & Services
LiFang Chen, Jiawei Li, Hongze Ge
Summary: In this paper, a network model called TBUnet is proposed for medical image segmentation. TBUnet extracts high frequency, low frequency, and boundary information through three branches, and uses a fusion layer and a feature enhancement module to combine and emphasize features. Experiments demonstrate that TBUnet achieves excellent segmentation performance and generalization capability on different datasets.
JOURNAL OF MEDICAL SYSTEMS
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Po-hung Wu, Mariajose Bedoya, Jim White, Christopher L. Brace
Summary: An automated ablation segmentation technique was developed for serial low-dose, noisy CT or CECT imaging, which can accurately assess the ablation geometry. Fuzzy c-means clustering may aid automatic segmentation of ablation zones without prior information or user input, making it more potential to assess treatments intra-procedurally.
Article
Ecology
Anamika Banwari, Rakesh Chandra Joshi, Namita Sengar, Malay Kishore Dutta
Summary: In this proposed algorithm, a computer vision-based technique is developed to predict the freshness level of fish from its image. By extracting features from the region of interest (fish eye) and observing the degradation pattern of these features, the freshness level of the sample fish can be accurately determined. The proposed method achieves a high recognition accuracy and low computation time, making it efficient for real-world usage in the fish industry and market.
ECOLOGICAL INFORMATICS
(2022)
Article
Computer Science, Interdisciplinary Applications
Shuai Wang, Mingxia Liu, Jun Lian, Dinggang Shen
Summary: This study introduces a novel boundary coding network (BCnet) to learn discriminative representations of organ boundaries for segmentation in male pelvic CT images. Experimental results show that this method outperforms several state-of-the-art methods in terms of segmentation accuracy.
IEEE TRANSACTIONS ON MEDICAL IMAGING
(2021)
Article
Oncology
Sanne G. M. van Velzen, Steffen Bruns, Jelmer M. Wolterink, Tim Leiner, Max A. Viergever, Helena M. Verkooijen, Ivana Isgum
Summary: This study aims to develop and evaluate an automatic deep learning method for segmentation of cardiac chambers and large arteries, and localization of the 3 main coronary arteries in radiation therapy planning on computed tomography (CT). The study found that the developed method can automatically obtain accurate estimates of planned radiation dose and dosimetric parameters for the cardiac chambers, large arteries, and coronary arteries.
INTERNATIONAL JOURNAL OF RADIATION ONCOLOGY BIOLOGY PHYSICS
(2022)
Article
Respiratory System
Anton Schreuder, Colin Jacobs, Nikolas Lessmann, Mireille J. M. Broeders, Mario Silva, Ivana Isgum, Pim A. de Jong, Michel M. van den Heuvel, Nicola Sverzellati, Mathias Prokop, Ugo Pastorino, Cornelia M. Schaefer-Prokop, Bram van Ginneken
Summary: This study developed and validated predictive models of competing death risk using CT information to identify participants with low lung cancer risk but high death risk. By considering lung cancer incidence, a subgroup with higher CD risk was identified, improving screening efficiency and enabling multidisciplinary treatment pathways.
EUROPEAN RESPIRATORY JOURNAL
(2022)
Article
Computer Science, Artificial Intelligence
Jorg Sander, Bob D. de Vos, Ivana Isgum
Summary: This paper proposes an unsupervised deep learning semantic interpolation approach for synthesizing high-resolution medical images from low-resolution examples. The method utilizes the latent space generated by autoencoders and employs convex combination to achieve semantically smooth interpolation. Evaluation on multiple medical image datasets shows that the proposed method outperforms traditional methods in terms of structural similarity and peak signal-to-noise ratio.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Computer Science, Artificial Intelligence
Louis D. van Harten, Catharina S. de Jonge, Kim J. Beek, Jaap Stoker, Ivana Isgum
Summary: Cine-MRI of the abdomen is a valuable technique for assessing small intestinal motility, but the potential of 3D cine-MRI has been underexplored. Existing image analysis tools have limitations in studying intestinal motility as 3D structures, and deep learning-based segmentation of the small intestine performs poorly due to the anatomical variations among patients. This study proposes a multi-task method that tracks individual segments of the small intestine and performs segmentation using a stochastic tracker and a CNN-based orientation classifier, conditioned on the locations of intestinal centerlines.
MEDICAL IMAGE ANALYSIS
(2022)
Article
Oncology
Sanne G. M. van Velzen, Roxanne Gal, Arco J. Teske, Femke van der Leij, Desiree H. J. G. van den Bongard, Max A. Viergever, Helena M. Verkooijen, Ivana Isgum
Summary: The study aimed to investigate whether the planned dose for cardiac structures is associated with the risk of heart disease in breast cancer patients receiving radiation therapy, and whether this association is influenced by coronary artery calcification. The results showed that radiation exposure to cardiac structures is associated with an increased risk of heart disease, particularly in patients with breast cancer and coronary artery calcification.
INTERNATIONAL JOURNAL OF RADIATION ONCOLOGY BIOLOGY PHYSICS
(2022)
Article
Health Care Sciences & Services
Annelotte Vos, Aryan Vink, Remko Kockelkoren, Richard A. P. Takx, Csilla Celeng, Willem P. T. M. Mali, Ivana Isgum, Ronald L. A. W. Bleys, Pim A. de Jong
Summary: This study investigated the detection and characterization of arterial calcifications in leg arteries using radiography and computed tomography (CT) and compared them with histology. The results showed that both radiography and CT could detect the majority of calcifications, but missed some mild calcifications. There was a moderate agreement between radiography/CT and histology in determining the location of calcifications in the intima or media of the arteries.
JOURNAL OF PERSONALIZED MEDICINE
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
Sanne G. M. van Velzen, Bob D. de Vos, Julia M. H. Noothout, Helena M. Verkooijen, Max A. Viergever, Ivana Isgum
Summary: This study proposes a CAC quantification method that does not require a threshold for segmentation, using a generative adversarial network (GAN) for image decomposition. The method improves the interscan reproducibility of CAC scoring compared to clinical calcium scoring.
JOURNAL OF MEDICAL IMAGING
(2022)
Article
Cardiac & Cardiovascular Systems
Mimount Bourfiss, Jorg Sander, Bob D. de Vos, Anneline S. J. M. Te Riele, Folkert W. Asselbergs, Ivana Isgum, Birgitta K. Velthuis
Summary: This study applies automatic deep learning-based segmentation for right and left ventricular CMR assessment and combines it with manual correction to accurately classify subjects suspected of ARVC according to CMR TFC.
CLINICAL RESEARCH IN CARDIOLOGY
(2023)
Article
Radiology, Nuclear Medicine & Medical Imaging
Annelotte Vos, Ignas B. Houben, Csilla Celeng, Richard A. P. Takx, Ivana Isgum, Willem P. T. M. Mali, Aryan Vink, Pim A. de Jong
Summary: This study aimed to validate the detectability and location of aortic calcification detected by computed tomography (CT) through histology. The results showed that CT reliably determined the presence and annularity of calcifications, which were mainly located in the intimal layer of the abdominal aorta.
EUROPEAN JOURNAL OF RADIOLOGY
(2023)
Proceedings Paper
Engineering, Biomedical
Jingnan Jia, Marius Staring, Irene Hernandez-Giron, Lucia J. M. Kroft, Anne A. Schouffoer, Berend C. Stoel
Summary: In this study, an automatic scoring framework using deep regression neural networks was proposed for visually scoring lung involvement in systemic sclerosis (SSc) from CT scans. The framework achieved competitive performance and has the potential to be an objective alternative for visual scoring of SSc in CT thorax studies.
MEDICAL IMAGING 2022: COMPUTER-AIDED DIAGNOSIS
(2022)
Proceedings Paper
Engineering, Biomedical
Yichao Li, Mohamed S. Elmandy, Michael S. Lew, Marius Staring
Summary: This study demonstrates the effectiveness of semi-supervised learning (SSL) in improving the performance of deep supervised models in the medical domain, even in high data regimes. By applying Stochastic Weight Averaging (SWA) technique, our model achieved significant improvements in prostate, seminal vesicles, rectum, and bladder detection compared to the supervised baseline in a prostate CT dataset.
MEDICAL IMAGING 2022: COMPUTER-AIDED DIAGNOSIS
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Zhiwei Zhai, Yining Wang, Bob D. de Vos, Julia M. H. Noothout, Nils Planken, Ivana Isgum
Summary: This study presents a computer algorithm trained with synthesized data to improve the detection and segmentation of coronary artery plaques.
MEDICAL IMAGING 2022: IMAGE PROCESSING
(2022)
Proceedings Paper
Computer Science, Interdisciplinary Applications
Prerak P. Mody, Nicolas Chaves-de-Plaza, Klaus Hildebrandt, Rene van Egmond, Huib de Ridder, Marius Staring
Summary: This study investigates the application of Bayesian models DropOut and FlipOut in automated contouring, and evaluates their performance using quantitative and qualitative metrics. The results show that Bayesian models have low uncertainty in accurate regions and high uncertainty in inaccurate regions, which helps to improve the quality assessment process.
MEDICAL IMAGING 2022: IMAGE PROCESSING
(2022)
Article
Radiology, Nuclear Medicine & Medical Imaging
G. Litjens, C. J. H. M. van Laarhoven, M. Prokop, E. J. M. van Geenen, J. J. Hermans
Summary: This retrospective study analyzed the effectiveness of routine CECT in staging and determining resectability of duodenal adenocarcinoma. The results showed that CECT underestimates T-stage and N-stage, and M-stage is often unclear, resulting in futile surgery for some patients.
ABDOMINAL RADIOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Olaf M. Neve, Yunjie Chen, Qian Tao, Stephan R. Romeijn, Nick P. de Boer, Willem Grootjans, Mark C. Kruit, Boudewijn P. F. Lelieveldt, Jeroen C. Jansen, Erik F. Hensen, Berit M. Verbist, Marius Staring
Summary: This study developed an automated method for measuring vestibular schwannoma on contrast-enhanced T1- and T2-weighted MRI scans. The convolutional neural network model achieved accurate delineation of the tumor and differentiation between intrameatal and extrameatal tumor parts.
RADIOLOGY-ARTIFICIAL INTELLIGENCE
(2022)