Article
Multidisciplinary Sciences
Xiyuan Su, Changqing Cao, Xiaodong Zeng, Zhejun Feng, Jingshi Shen, Xu Yan, Zengyan Wu
Summary: This study proposes a fault diagnosis method based on deep belief networks and restricted Boltzmann machines, combined with grey wolf optimization algorithm, to improve the accuracy and efficiency of analog circuit fault diagnosis.
SCIENTIFIC REPORTS
(2021)
Article
Chemistry, Analytical
Guangyi Chen, Adam Krzyzak
Summary: In this paper, a novel method for 2D pattern recognition is proposed, which utilizes the log-polar transform, dual-tree complex wavelet transform (DTCWT), and 2D fast Fourier transform (FFT2) to extract features. The new method achieves translation, rotation, and scaling invariance for input 2D pattern images in a multiresolution manner, which is crucial for invariant pattern recognition. Experimental results on a printed Chinese character dataset and a 2D aircraft dataset demonstrate that the new method outperforms two existing methods in most testing cases involving rotations, scaling, and various noise levels in the input pattern images.
Article
Computer Science, Information Systems
Jisha Anu Jose, C. Sathish Kumar, S. Sureshkumar
Summary: Tuna fish is commercially important and its classification plays a crucial role in the fishing industry. This study presents an automated tuna classification system using textural macro features, achieving high accuracy and performance.
JOURNAL OF KING SAUD UNIVERSITY-COMPUTER AND INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Kui Liu, Yiping Guo, Benyue Su
Summary: This paper proposes an image denoising network based on dual-tree complex wavelet for the recovery of complex texture regions. The network combines spatial and transform domains using dual-tree complex wavelet transform (DTCWT) to capture structured features and time-frequency localized features. The network utilizes a Subband Information Sharing Unit (SISU) to enhance recovery in hard scenes and rectified linear units and exponential linear units to match properties of elements in different domains. Experimental results demonstrate the network's powerful recovery capability and competitive performance in non-blind/blind image denoising.
NEURAL PROCESSING LETTERS
(2023)
Article
Computer Science, Artificial Intelligence
Nicolas Couellan
Summary: This study focuses on the robustness of deep neural networks, proposing a simple concentration inequality to study the probability of network output deviating from its nominal value and using network conditions to regularize the loss function. Empirical evaluation shows that the proposed method accurately estimates the observed robustness.
Article
Computer Science, Artificial Intelligence
Mohamed Yamni, Hicham Karmouni, Mhamed Sayyouri, Hassan Qjidaa
Summary: The proposed audio watermarking scheme based on a novel hybrid transform combines DTCWT and FrCMT to achieve high robustness and security by embedding watermark bits in coefficients.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Engineering, Mechanical
Huan Wang, Zhiliang Liu, Dandan Peng, Ming J. Zuo
Summary: This paper proposes a multilayer wavelet attention convolutional neural network (MWA-CNN) for noise-robust machinery fault diagnosis. The framework aims to learn discriminative fault features from the wavelet domain, which allows the model to obtain better interpretability and superior performance than conventional time-domain-based CNNs. Experiments on high-speed aeronautical bearing and motor bearing datasets prove that the proposed method has excellent fault diagnosis ability and noise robustness, and the visual analysis of the attention mechanism contributes to the interpretability of CNN in the field of fault diagnosis.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2023)
Article
Engineering, Multidisciplinary
Xin Zhang, Zhongqiang Zhang, Jiaxu Wang, Zhiwen Liu, Lei Wang
Summary: The paper proposes a novel method called Reweighted-Kurtogram with sub-bands rearranged and ensemble dual-tree complex wavelet packet transform (SRE-DTCWPT) to improve the performance of the Fast Kurtogram (FK) in bearing fault diagnosis. The method involves a set of envelope analysis approaches and introduces a new robust evaluating indicator called reweighted kurtosis. Experimental results show that the proposed method has high potentials for extracting bearing diagnostic information from complex vibration signals.
STRUCTURAL HEALTH MONITORING-AN INTERNATIONAL JOURNAL
(2022)
Article
Computer Science, Artificial Intelligence
K. Reddy Madhavi, Padmavathi Kora, L. Venkateswara Reddy, J. Avanija, K. L. S. Soujanya, Prabhakar Telagarapu
Summary: This study uses deep learning and dual-tree wavelet transform to detect abnormal heartbeats, and the results show that the deep learning architecture has a high accuracy in heartbeat classification. This method can help doctors accurately identify diseases and reduce misdiagnosis.
Article
Computer Science, Interdisciplinary Applications
J. Lin, W. Zhou, X. Z. Cui, H. P. Hong
Summary: This study presents a framework to analyze and simulate nonhomogeneous non-Gaussian corrosion fields on buried in-service pipelines using continuous and discrete wavelet transforms. The framework incorporates the measured corrosion field and utilizes different wavelet transforms and an iterative algorithm to generate synthetic corrosion fields. The results indicate that the proposed framework effectively captures the probabilistic characteristics of the measured corrosion field and can be used for pipeline corrosion management.
COMPUTERS & STRUCTURES
(2023)
Article
Computer Science, Information Systems
Ahmadreza Sezavar, Randa Atta, Mohammad Ghanbari
Summary: In this paper, a person identification framework based on smartphone-acquired gait signals is proposed, which combines convolutional neural network (CNN) and dual-tree complex wavelet transform (DTCWT) to achieve higher recognition accuracy. Experimental results demonstrate the superiority of the proposed method over the state-of-the-art models.
MULTIMEDIA SYSTEMS
(2022)
Article
Chemistry, Physical
M. A. Rubio, D. G. Sanchez, P. Gazdzicki, K. A. Friedrich, A. Urquia
Summary: Early diagnosis of fuel cell failure modes using local electrochemical noise analysis is proposed in this study. The method is suitable for real-time diagnosis, achieves a high identification rate, and can be implemented using lightweight and inexpensive hardware.
JOURNAL OF POWER SOURCES
(2022)
Article
Computer Science, Information Systems
Sanjay Kumar Maurya, Ravindra Kumar Singh
Summary: This paper introduces a Super-Resolution (SR) technique based on constructing DT-CWT coefficients for larger scale images by predicting phase and estimating magnitude for each subband at a finer level. The proposed method has been shown to outperform conventional and state-of-the-art techniques in terms of quantitative and visual results when applied to various types of images.
MULTIMEDIA TOOLS AND APPLICATIONS
(2021)
Article
Optics
Randa Atta, Mohammad Ghanbari
Summary: This paper proposes an image steganography method based on dual tree complex wavelet transform (DT-CWT) to increase capacity and maintain image quality. By utilizing the human visual system's insensitivity to edges, more secret bits can be hidden, making the method superior in terms of capacity and image fidelity compared to existing methods.
Article
Biology
Ela Kaplan, Sengul Dogan, Turker Tuncer, Mehmet Baygin, Erman Altunisik
Summary: In this study, a new automatic AD detection model called LPQNet was proposed, demonstrating high classification accuracy on three different image datasets and showing superiority over other detection models. Additionally, LPQNet can be used to develop a new generation intelligent AD detection application for MRI and CT devices.
COMPUTERS IN BIOLOGY AND MEDICINE
(2021)
Article
Energy & Fuels
Ji-In Kim, Hui-Seon Gang, Jae-Young Pyun, Goo-Rak Kwon
Summary: This study proposes a vision-based indoor positioning method using QR code, accurately measuring a user's position by determining the current position of a smartphone device based on a QR code recognized with a smartphone camera.
Article
Computer Science, Artificial Intelligence
Tobias Ross, Annika Reinke, Peter M. Full, Martin Wagner, Hannes Kenngott, Martin Apitz, Hellena Hempe, Diana Mindroc-Filimon, Patrick Scholz, Thuy Nuong Tran, Pierangela Bruno, Pablo Arbelaez, Gui-Bin Bian, Sebastian Bodenstedt, Jon Lindstrom Bolmgren, Laura Bravo-Sanchez, Hua-Bin Chen, Cristina Gonzalez, Dong Guo, Pal Halvorsen, Pheng-Ann Heng, Enes Hosgor, Zeng-Guang Hou, Fabian Isensee, Debesh Jha, Tingting Jiang, Yueming Jin, Kadir Kirtac, Sabrina Kletz, Stefan Leger, Zhixuan Li, Klaus H. Maier-Hein, Zhen-Liang Ni, Michael A. Riegler, Klaus Schoeffmann, Ruohua Shi, Stefanie Speidel, Michael Stenzel, Isabell Twick, Gutai Wang, Jiacheng Wang, Liansheng Wang, Lu Wang, Yujie Zhang, Yan-Jie Zhou, Lei Zhu, Manuel Wiesenfarth, Annette Kopp-Schneider, Beat P. Mueller-Stich, Lena Maier-Hein
Summary: Robustness and generalization capabilities are crucial for medical instrument segmentation algorithms in endoscopic image processing. The ROBUST-MIS challenge evaluated algorithm performance on challenging images and across different surgical interventions, highlighting the importance of these capabilities.
MEDICAL IMAGE ANALYSIS
(2021)
Article
Computer Science, Information Systems
Debesh Jha, Pia H. Smedsrud, Dag Johansen, Thomas de Lange, Havard D. Johansen, Pal Halvorsen, Michael A. Riegler
Summary: In this paper, the authors propose using Conditional Random Field and Test-Time Augmentation to further improve the prediction performance of the ResUNet++ architecture, validated on multiple public datasets, especially showing good results for the detection of flat or sessile polyps.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2021)
Article
Multidisciplinary Sciences
Pia H. Smedsrud, Vajira Thambawita, Steven A. Hicks, Henrik Gjestang, Oda Olsen Nedrejord, Espen Naess, Hanna Borgli, Debesh Jha, Tor Jan Derek Berstad, Sigrun L. Eskeland, Mathias Lux, Havard Espeland, Andreas Petlund, Duc Tien Dang Nguyen, Enrique Garcia-Ceja, Dag Johansen, Peter T. Schmidt, Ervin Toth, Hugo L. Hammer, Thomas de Lange, Michael A. Riegler, Pal Halvorsen
Summary: AI is expected to improve anomaly detection and reduce manual labor in VCE technology, and the Kvasir-Capsule dataset can assist in developing better algorithms to fully utilize the potential of VCE technology.
Article
Biology
Rabindra Khadka, Debesh Jha, Steven Hicks, Vajira Thambawita, Michael A. Riegler, Sharib Ali, Pal Halvorsen
Summary: Traditional supervised deep learning methods often fail to generalize on unseen datasets in medical imaging. Few-shot learning approaches can minimize the need for a large number of training samples and be used for modeling new datasets, showing great potential.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Information Systems
Abhishek Srivastava, Debesh Jha, Sukalpa Chanda, Umapada Pal, Havard Johansen, Dag Johansen, Michael Riegler, Sharib Ali, Pal Halvorsen
Summary: Methods based on convolutional neural networks have improved biomedical image segmentation, but most cannot efficiently handle diverse object sizes or small, biased datasets. This paper introduces the MSRF-Net architecture, specifically designed for medical image segmentation, which uses DSDF blocks to exchange multi-scale features and improve information flow, allowing for accurate segmentation. Extensive experiments show that MSRF-Net outperforms current methods on publicly available datasets, achieving high Dice Coefficients on various biomedical datasets.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Mathematics
Uttam Khatri, Goo-Rak Kwon
Summary: This study proposes a new diagnosis scheme by constructing high-order dynamic functional networks for the diagnosis of Alzheimer's disease. The experimental results demonstrate the excellent performance of the proposed scheme in terms of diagnosis accuracy, validating its effectiveness in clinical value.
Article
Mathematics
Rukesh Prajapati, Goo-Rak Kwon
Summary: Proper analysis of changes in brain structure is crucial for diagnosing brain disorders accurately. In this study, a novel architecture combining three parallel UNets and a residual network is proposed to improve upon baseline methods. By using three consecutive images as input, compressing and decompressing them individually, and enhancing image features with skip connections, the proposed architecture outperforms single conventional UNet and other UNet variants in segmentation accuracy.
Article
Mathematics
Ramesh Kumar Lama, Ji-In Kim, Goo-Rak Kwon
Summary: This study classified Alzheimer's disease and mild cognitive impairment from healthy controls by constructing brain networks from functional magnetic resonance images and using different feature selection methods and classifiers. Experimental results showed that using the LASSO feature selection method in large networks and the FSAL feature selection technique in small networks improved classification accuracy.
Article
Chemistry, Multidisciplinary
Rukesh Prajapati, Goo-Rak Kwon
Summary: This study proposes a method to improve the accuracy of brain segmentation by using generative adversarial networks and a novel corrective algorithm. The generator weights are updated multiple times and the dissimilarity is calculated to minimize false predictions.
APPLIED SCIENCES-BASEL
(2022)
Article
Mathematics
Uttam Khatri, Ji-In Kim, Goo-Rak Kwon
Summary: We aim to identify patients with mild cognitive impairment (MCI) at risk of progressing to Alzheimer's disease (AD) by using high-dimensional-data resting state functional magnetic resonance imaging (rs-fMRI) brain networks and gene expression. Integrating genetic traits with brain imaging for clinical examination is limited. However, genetic markers can effectively diagnose healthy controls (HCs) and the two phases of MCI (convertible and stable MCI) as well as AD.
Article
Computer Science, Artificial Intelligence
Nikhil Kumar Tomar, Debesh Jha, Michael A. Riegler, Havard D. Johansen, Dag Johansen, Jens Rittscher, Pal Halvorsen, Sharib Ali
Summary: This paper introduces a novel architecture called feedback attention network (FANet) that leverages the information of each training epoch to prune the prediction maps of the subsequent epochs and rectify the predictions iteratively during the test time. Experimental results demonstrate that FANet provides significant improvement on segmentation metrics tested on various biomedical imaging datasets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Medicine, General & Internal
Ramesh Kumar Lama, Goo-Rak Kwon
Summary: This study investigates the clustering structure of functional connectivity in different brain networks in healthy, AD, and prodromal stage subjects. The findings suggest that the disruption of connectivity in certain networks, such as the sensory motor network, dorsal attention network, salience network, default mode network, and cerebral network, may serve as potential biomarkers for distinguishing AD and MCI from healthy individuals. These alterations in functional connectivity within these networks may contribute to the cognitive deficits observed in AD and MCI.
Article
Computer Science, Information Systems
Fazal Ur Rehman Faisal, Goo-Rak Kwon
Summary: This study aims to develop a deep learning method for extracting valuable Alzheimer's disease biomarkers from structural magnetic resonance imaging (sMRI) and classifying brain images into different groups. The proposed method shows superior results compared to existing methods, with reduced parameters and computation complexity.
Proceedings Paper
Computer Science, Interdisciplinary Applications
Debesh Jha, Nikhil Kumar Tomar, Sharib Ali, Michael A. Riegler, Havard D. Johansen, Dag Johansen, Thomas de Lange, Pal Halvorsen
Summary: This research introduces a novel architecture NanoNet for endoscopic image segmentation, providing real-time performance and improved segmentation accuracy. By evaluating with various datasets, the effectiveness of this method is validated. Compared to traditional deep learning approaches, this architecture has a smaller model size and higher performance.
2021 IEEE 34TH INTERNATIONAL SYMPOSIUM ON COMPUTER-BASED MEDICAL SYSTEMS (CBMS)
(2021)
