Article
Computer Science, Artificial Intelligence
M. Tanveer, A. Tiwari, R. Choudhary, M. A. Ganaie
Summary: This study proposes a novel large scale pinball twin support vector machine (LPTWSVM) to address the limitations of the twin support vector machines (TWSVMs), using a unique pinball loss function and improving model performance by eliminating matrix inversion calculation and minimizing structural risk.
Article
Computer Science, Artificial Intelligence
Chun-Na Li, Yuan-Hai Shao, Huajun Wang, Yu-Ting Zhao, Naihua Xiu, Nai-Yang Deng
Summary: This paper investigates the general forms and characteristics of nonparallel support vector machines (NSVMs) and categorizes them into two types. It reveals the advantages and defects of different types and points out the inconsistency problems. Based on this observation, a novel max-min distance-based NSVM is proposed with desired consistency. The proposed NSVM has the consistency of training and test and the consistency of metric, and it assigns each sample an ascertained loss.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Xiaoming Wang, Shitong Wang, Zengxi Huang, Yajun Du
Summary: This paper introduces a novel method called sparse support vector machine guided by radius-margin bound (RMB-SSVM) to efficiently condense the basis vectors in support vector machines. By selecting basis vectors and learning corresponding coefficients with a criterion related to SVM's generalization ability, the RMB-SSVM model can yield better performance.
APPLIED SOFT COMPUTING
(2021)
Article
Computer Science, Artificial Intelligence
Sambhav Jain, Reshma Rastogi
Summary: This paper proposes Parametric non-parallel support vector machines for binary pattern classification. The model brings noise resilience and sparsity by intelligently redesigning the Support vector machine optimization. The experimental results validate its scalability for large scale problems.
Article
Computer Science, Information Systems
Sebastian Maldonado, Julio Lopez, Carla Vairetti
Summary: The predictive performance of classification methods relies heavily on the nature of the environment and dataset shift issue. A novel Fuzzy Support Vector Machine strategy is proposed in this paper to improve performance by redefining the loss function and applying aggregation operators to deal with dataset shift. Our methods outperform traditional classifiers in terms of out-of-time prediction using simulated and real-world dataset for credit scoring.
INFORMATION SCIENCES
(2021)
Article
Medicine, General & Internal
Deborah L. Harrington, Po-Ya Hsu, Rebecca J. Theilmann, Annemarie Angeles-Quinto, Ashley Robb-Swan, Sharon Nichols, Tao Song, Lu Le, Carl Rimmele, Scott Matthews, Kate A. Yurgil, Angela Drake, Zhengwei Ji, Jian Guo, Chung-Kuan Cheng, Roland R. Lee, Dewleen G. Baker, Mingxiong Huang
Summary: This study aimed to build a diagnostic model for blast-related mild traumatic brain injury (bmTBI) using machine learning based on diffusion tensor imaging (DTI) datasets. The study identified white-matter features that distinguish bmTBI from healthy controls (HC), and found that decreased radial diffusivity (RD), increased fractional anisotropy (FA), and axial/radial diffusivity ratio (AD/RD) were the most prominent features in bmTBI patients. The model achieved an accuracy rate of 89% in distinguishing bmTBI from HC.
Article
Optics
Rendong Ji, Zhezhen Jiang, Xiaoyan Wang, Yue Han, Haiyi Bian, Yudong Yang, Liyun Zhuang, Yulin Zhang
Summary: The study detects captan residues in apple juice using fluorescence spectrometry and predicts their content level using a genetic algorithm and support vector machines. The results show that this method has high accuracy and efficiency, making it suitable for rapid detection of captan residues in apple juice.
Article
Computer Science, Artificial Intelligence
Wangyong Lv, Tingting Li, Huali Ren, Shijing Zeng, Jiao Zhou
Summary: The IDH-MSVM algorithm adjusts the distance between hyperplanes and classical margins to handle multiclassification problems more flexibly. Experimental results on UCI standard data sets show that this method achieves better classification accuracy for multiclass data compared to other algorithms.
INTERNATIONAL JOURNAL OF INTELLIGENT SYSTEMS
(2022)
Article
Engineering, Biomedical
Sinem Burcu Erdogan, Gulnaz Yukselen, Mustafa Mert Yegul, Ruhi Usanmaz, Engin Kiran, Orhan Derman, Ata Akin
Summary: This study developed a novel fNIRS-based classification approach combining clinical, behavioral, and neurophysiological features to better identify impulsive adolescents. Support vector machines and artificial neural networks achieved diagnostic accuracies over 90% when trained with multi-domain features.
JOURNAL OF NEURAL ENGINEERING
(2021)
Article
Automation & Control Systems
Ai-Min Yang, Yang Han, Chen-Shuai Liu, Jian-Hui Wu, Dian-Bo Hua
Summary: The secondary application of medical big data is becoming increasingly popular in healthcare services and clinical research. Studying cancer recurrence can provide effective clinical intervention means for patients. This article establishes a cancer recurrence prediction model using the improved TSVR algorithm, with a prediction accuracy of over 91% for various cancers.
IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS
(2021)
Article
Computer Science, Information Systems
Yange Chen, Qinyu Mao, Baocang Wang, Pu Duan, Benyu Zhang, Zhiyong Hong
Summary: This paper presents a privacy-preserving medical diagnosis scheme based on multi-class support vector machines. By utilizing distributed public key encryption system and secure computing protocol, the scheme can handle different types of data while protecting the privacy of user data.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2022)
Article
Mathematics
Roberto Barcenas, Maria Gonzalez-Lima, Joaquin Ortega, Adolfo Quiroz
Summary: The effectiveness of subsampling methods in reducing the required instances in the training stage of using support vector machines (SVMs) for classification in big data scenarios is explored in this paper with theoretical results. The main theorem states that, under certain conditions, a feasible solution can be found for the SVM problem using a randomly chosen subsample, which can be as close as desired to the classifier trained with the complete dataset in terms of classification error. Additionally, a new subsampling method called importance sampling and bagging is proposed, which provides a faster solution to the SVM problem without significant loss in accuracy compared to existing techniques.
Article
Computer Science, Information Systems
Jing Wang, Libing Wu, Huaqun Wang, Kim-Kwang Raymond Choo, Debiao He
Summary: The proposed EPoSVM scheme is designed for IoMT deployment, protecting training data privacy and ensuring the security of the trained SVM model through secure computation protocols. Security analysis confirms that the protocols and EPoSVM satisfy both security and privacy protection requirements, with performance evaluation on real-world disease data sets showing its efficiency and effectiveness in achieving the same classification accuracy as a general SVM.
IEEE INTERNET OF THINGS JOURNAL
(2021)
Article
Neurosciences
Zhibo Wan, Youqiang Dong, Zengchen Yu, Haibin Lv, Zhihan Lv
Summary: The study investigates the feature recognition, diagnosis, and forecasting performances of Semi-Supervised Support Vector Machines (S3VMs) for brain image fusion Digital Twins (DTs), constructing a diagnostic and predictive model based on this approach. Results show that the model excels in accuracy and acceleration efficiency.
FRONTIERS IN NEUROSCIENCE
(2021)
Article
Biochemical Research Methods
Sandryne David, Trang Tran, Frederick Dallaire, Guillaume Sheehy, Feryel Azzi, Dominique Trudel, Francine Tremblay, Atilla Omeroglu, Frederic Leblond, Sarkis Meterissian
Summary: As many as 60% of patients with early stage breast cancer undergo breast-conserving surgery, but a significant number of them need a second surgery due to incomplete resection of the lesions. This study developed a technique using Raman spectroscopy and machine learning to detect invasive breast cancer in surgically resected breast specimens. The results showed high sensitivity and specificity in detecting cancer cells on the margins of the specimens.
JOURNAL OF BIOMEDICAL OPTICS
(2023)
Article
Clinical Neurology
Samuel Hall, Ashraf Abouharb, Ian Anderson, Andrew Bacon, Anuj Bahl, Howard Brydon, Graham Dow, Ioannis Fouyas, James Galea, Anthony Ghosh, Nihal Gurusinghe, Mahmoud Kamel, Pawan Minhas, Patrick Mitchell, David Mowle, Nitin Mukerji, Ramesh Nair, John Norris, Hiren Patel, Jash Patel, Krunal Patel, Jerome St George, Mario Teo, Ahmed Toma, Rikin Trivedi, Chris Uff, Anna Visca, Daniel C. Walsh, Edward White, Peter Whitfield, Diederik Bulters
Summary: This study investigated the current practices of surveillance for unruptured intracranial aneurysms (UIA) in the United Kingdom. The results showed significant heterogeneity in surveillance practices between different units, with lack of uniform policies and standards. This study contributes to better understanding the variation in practice among units and provides a framework for further research on aneurysm growth.
BRITISH JOURNAL OF NEUROSURGERY
(2023)
Article
Food Science & Technology
Samuel Ortega, Stein-Kato Lindberg, Stein Harris Olsen, Kathryn E. Anderssen, Karsten Heia
Summary: Mushy Halibut Syndrome (MHS) is a condition characterized by abnormal flesh texture in Greenland halibut, resulting in poor meat quality. This study proposes the use of hyperspectral imaging for MHS detection, aiming to estimate the chemical composition of affected samples and classify specimens using PLS-DA. The results show accurate prediction of fat content using spectral data, while the prediction of water content was less accurate. However, MHS detection using PLS-DA achieved high precision and recall rates for hyperspectral images, suggesting hyperspectral imaging as a suitable technology for early screening of MHS.
LWT-FOOD SCIENCE AND TECHNOLOGY
(2023)
Article
Optics
Laura Quintana-Quintana, Samuel Ortega, Himar Fabelo, Francisco J. Balea-Fernandez, Gustavo M. Callico
Summary: Hyperspectral imaging improves cancer diagnosis by automatically classifying cells, but achieving homogeneous focus in such images is challenging. This study aims to quantify the focus of hyperspectral images for image correction. State-of-the-art methods like Maximum Local Variation, Fast Image Sharpness block-based Method, and Local Phase Coherence algorithms showed the best correlation results with subjective scores of image focus.
Article
Chemistry, Analytical
Abian Hernandez-Guedes, Natalia Arteaga-Marrero, Enrique Villa, Gustavo M. Callico, Juan Ruiz-Alzola
Summary: Diabetes mellitus has a high global prevalence, and its long-term complication, diabetic foot ulcer (DFU), affects roughly 6.3% of the population and has a lifetime incidence of up to 34%. Infrared thermograms can be used to monitor the risk of DFU development by detecting abnormal foot patterns in diabetic patients. This study utilized the INAOE dataset along with additional data to extract features and classify DFU subjects, achieving improved performance compared to previous state-of-the-art features.
Article
Chemistry, Analytical
Carlos Urbina Ortega, Eduardo Quevedo Gutierrez, Laura Quintana, Samuel Ortega, Himar Fabelo, Lucana Santos Falcon, Gustavo Marrero Callico
Summary: This work proposes an approach to enhance the spatial resolution of hyperspectral histology samples using super-resolution. By taking multiple images of the same scene in a controlled environment with sub-pixel shifts between them, the proposed algorithm can effectively improve the spatial resolution of the sensor while preserving the spectral signature of the pixels, competing well with other state-of-the-art super-resolution techniques, and opening up possibilities for real-time applications.
Article
Computer Science, Information Systems
Antonio J. Rodriguez-Almeida, Himar Fabelo, Samuel Ortega, Alejandro Deniz, Francisco J. Balea-Fernandez, Eduardo Quevedo, Cristina Soguero-Ruiz, Ana M. Wagner, Gustavo M. Callico
Summary: The increasing prevalence of chronic non-communicable diseases has made it important to develop tools for improving their management. Artificial Intelligence algorithms have been successful in early diagnosis, prediction, and analysis in the medical field. However, there are challenges in dealing with medical data, such as the lack of high-quality datasets and the need to maintain patient privacy. Synthetic data generation techniques have emerged as a possible solution. This study develops a framework based on synthetic data generation algorithms and tests it using eight medical datasets. The preservation of data integrity and the maintenance of classification performance are evaluated using statistical metrics and machine learning classifiers.
IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS
(2023)
Article
Chemistry, Multidisciplinary
Clara Garcia-Vicente, David Chushig-Muzo, Inmaculada Mora-Jimenez, Himar Fabelo, Inger Torhild Gram, Maja-Lisa Lochen, Conceicao Granja, Cristina Soguero-Ruiz
Summary: This paper investigates the application of oversampling methods in generating synthetic categorical clinical data to improve the predictive performance of machine learning models for identifying risk factors and building models for cardiovascular disease prediction. Experimental results show that GAN-based models can generate high-quality synthetic data and improve the predictive performance of the models. Combining oversampling strategies with linear and nonlinear supervised ML methods further enhances the performance, providing insights into risk factors and facilitating CVD prediction.
APPLIED SCIENCES-BASEL
(2023)
Proceedings Paper
Computer Science, Information Systems
Clara Garcia-Vicente, David Chushig-Muzo, Inmaculada Mora-Jimenez, Himar Fabelo, Inger Torhild Gram, Maja-Lisa Lochen, Conceicao Granja, Cristina Soguero-Ruiz
Summary: Noncommunicable diseases, particularly cardiovascular diseases, pose significant health threats in society. Early detection and prevention are crucial for reducing the global burden. This study applies oversampling techniques to improve the identification of risk factors associated with cardiovascular diseases. The results show that oversampling and feature selection techniques help enhance the accuracy of cardiovascular disease prediction.
HETEROGENEOUS DATA MANAGEMENT, POLYSTORES, AND ANALYTICS FOR HEALTHCARE, DMAH 2022
(2022)
Article
Clinical Neurology
Isaac Josh Abecassis, R. Michael Meyer, Michael R. Levitt, Jason P. Sheehan, Ching-Jen Chen, Bradley A. Gross, Ashley Lockerman, W. Christopher Fox, Waleed Brinjikji, Giuseppe Lanzino, Robert M. Starke, Stephanie H. Chen, Adriaan R. E. Potgieser, J. Marc C. van Dijk, Andrew Durnford, Diederik Bulters, Junichiro Satomi, Yoshiteru Tada, Amanda Kwasnicki, Sepideh Amin-Hanjani, Ali Alaraj, Edgar A. Samaniego, Minako Hayakawa, Colin P. Derdeyn, Ethan Winkler, Adib Abla, Pui Man Rosalind Lai, Rose Du, Ridhima Guniganti, Akash P. Kansagra, Gregory J. Zipfel, Louis J. Kim
Summary: This study aims to understand the natural history, rate of spontaneous regression, and ideal treatment regimen of patients with intracranial dural arteriovenous fistulas (dAVFs) and concurrent intracranial aneurysms. The results show that patients with dAVFs have a similar risk of harboring an intracranial aneurysm unrelated to the dAVF (5.3%) compared with the general population (approximately 2%-5%), and only 50% of these aneurysms are intradural.
JOURNAL OF NEUROSURGERY
(2022)