Review
Biochemical Research Methods
Rufeng Li, Lixin Li, Yungang Xu, Juan Yang
Summary: The innovation of biotechnologies has accelerated the accumulation of omics data, leading to the era of 'big data'. Extracting valuable knowledge from omics data remains a challenging issue that requires innovative methods. The development and application of machine learning have greatly enhanced insights into biology and biomedicine, particularly in the field of precision medicine.
BRIEFINGS IN BIOINFORMATICS
(2022)
Review
Biochemistry & Molecular Biology
Romano Weiss, Sanaz Karimijafarbigloo, Dirk Roggenbuck, Stefan Roediger
Summary: Neural networks, also known as artificial neural networks, are crucial tools in deep-learning applications, and their popularity has soared since the early 2000s. This review focuses on the use of deep learning in biomedical data analysis, particularly in the analysis of biomarkers in bioimage data. The article also provides quantitative insights into the usage of network types in different scientific fields based on a data analysis of neural network publications.
Article
Computer Science, Information Systems
Israel Cruz-Vega, Hans Israel Morales-Lopez, Juan Manuel Ramirez-Cortes, Jose De Jesus Rangel-Magdaleno
Summary: Cataract is a disease that causes opacity in the ocular nucleus. Automatic classification and grading of the disease using computational intelligence methods require adequately-labeled databases. This study presents a new cataract database and compares different machine learning algorithms and deep learning structures for classification.
Review
Multidisciplinary Sciences
Yanan Sui, Huiling Yu, Chen Zhang, Yue Chen, Changqing Jiang, Luming Li
Summary: Deep brain-machine interfaces bridge machines and deep brain structures, offering promising treatments for neurological and psychiatric disorders. They differ from conventional brain-machine interfaces by enabling interactions with deep brain structures, sensing and modulating deep brain neural activities for function restoration and therapeutic improvements. This article provides an overview of deep brain recording and stimulation techniques serving as deep brain-machine interfaces, with a focus on deep brain stimulation and stereotactic electroencephalography, their technical trends, clinical applications, and potential for closed-loop systems in the treatment of neurological and psychiatric disorders.
NATIONAL SCIENCE REVIEW
(2022)
Review
Oncology
Jiaona Xu, Yuting Meng, Kefan Qiu, Win Topatana, Shijie Li, Chao Wei, Tianwen Chen, Mingyu Chen, Zhongxiang Ding, Guozhong Niu
Summary: Glioma, one of the most fatal primary brain tumors, poses difficulties in diagnosis and management. Medical imaging techniques and artificial intelligence play significant roles in the diagnosis and treatment of glioma.
FRONTIERS IN ONCOLOGY
(2022)
Article
Engineering, Biomedical
Eleonora Borda, Danashi Imani Medagoda, Marta Jole Ildelfonsa Airaghi Leccardi, Elodie Genevieve Zollinger, Diego Ghezzi
Summary: Off-stoichiometry thiol-ene-epoxy (OSTE+) thermosets have low gas permeability and little absorption of dissolved molecules, allowing for low-temperature dry bonding without surface treatments. They can be manufactured via UV polymerisation and have gained attention for rapid prototyping of microfluidic chips and as a novel material for neural implants due to their mechanical properties and compatibility with standard clean-room processes.
Article
Engineering, Biomedical
Nur Ahmadi, Timothy G. Constandinou, Christos-Savvas Bouganis
Summary: The study introduces a new method that achieves high decoding accuracy and long-term robustness in brain-machine interfaces. By using entire spiking activity as the input signal and coupling it with a deep learning-based decoding algorithm, experiments on non-human primates demonstrated consistently higher decoding performance over long-term recording sessions.
JOURNAL OF NEURAL ENGINEERING
(2021)
Article
Engineering, Biomedical
Joseph T. Costello, Samuel R. Nason-Tomaszewski, Hyochan An, Jungho Lee, Matthew J. Mender, Hisham Temmar, Dylan M. Wallace, Jongyup Lim, Matthew S. Willsey, Parag G. Patil, Taekwang Jang, Jamie D. Phillips, Hun-Seok Kim, David Blaauw, Cynthia A. Chestek
Summary: This study explores a transmission scheme for free-floating motes based on infrared, which reduces wireless data rate and system power consumption through pulse-interval modulation (PIM) communication. The results show that PIM has strong neural information transmission capability and matches the performance of traditional wired systems in real-time closed-loop BMI tests. Additionally, the PIM communication scheme is feasible in terms of power and enables high-channel-count high-performance BMIs.
JOURNAL OF NEURAL ENGINEERING
(2022)
Article
Materials Science, Multidisciplinary
Jack Maughan, Pedro J. Gouveia, Javier Gutierrez Gonzalez, Liam M. Leahy, Ian Woods, Cian O'Connor, Tara McGuire, James R. Garcia, Donagh G. O'Shea, Sarah F. McComish, Oran D. Kennedy, Maeve A. Caldwell, Adrian Dervan, Jonathan N. Coleman, Fergal J. O'Brien
Summary: This study reports the development of an electroconductive pristine graphene-based (pG) composite material for central nervous system applications, which balances conductivity, biocompatibility, and mechanical performance. The composite material showed no adverse effects on neuronal and glial cell growth, and enhanced neurite outgrowth and cellular morphology compared to controls. The versatility of the composite material was demonstrated through the fabrication of various conductive neural-interfacing structures.
APPLIED MATERIALS TODAY
(2022)
Review
Biochemical Research Methods
Sachin Kumar, Karan Veer
Summary: With the rapid development of high-volume and complex data analysis, machine learning has become a critical tool for classification and prediction. This study reviews the methods of machine learning and deep learning for the classification and prediction of biological signals and conducts a systematic review on their applications in different biomedical signals from 2015 to 2022. The findings show a clear shift towards deep learning techniques in the classification of biomedical signals compared to machine learning.
CURRENT BIOINFORMATICS
(2023)
Article
Engineering, Biomedical
Byron Llerena Zambrano, Aline F. Renz, Tobias Ruff, Samuel Lienemann, Klas Tybrandt, Janos Voros, Jaehong Lee
Summary: This article systematically describes the recent progress of implantable soft electronics based on various conductive nanocomposites, focusing on representative fabrication approaches and various in vivo applications. The mechanical mismatch between implanted devices and biological environments can induce damages in the body, especially for long-term applications, which stretchable electronics effectively improve.
ADVANCED HEALTHCARE MATERIALS
(2021)
Review
Computer Science, Information Systems
Shi Dong, Ping Wang, Khushnood Abbas
Summary: This paper primarily focuses on the development, classic models, latest applications, problems, and future research directions of deep learning in various fields such as speech processing, computer vision, natural language processing, and medical applications.
COMPUTER SCIENCE REVIEW
(2021)
Article
Neurosciences
Aritra Das, Nilanjana Nandi, Supratim Ray
Summary: Attention typically increases power in the gamma band and decreases power in the alpha band in brain signals. The power of steady-state visually evoked potential (SSVEP) also increases with attention. However, the effectiveness of these neural measures in capturing attentional modulation is unclear. A recent study in macaques found that attentional modulation was more prominent in the gamma band. To compare this with human EEG, the researchers conducted an experiment and found that attentional modulation was comparable for SSVEP and alpha, while non-foveal stimuli had minimal attentional modulation.
Article
Computer Science, Artificial Intelligence
Shams Forruque Ahmed, Md. Sakib Bin Alam, Maruf Hassan, Mahtabin Rodela Rozbu, Taoseef Ishtiak, Nazifa Rafa, M. Mofijur, A. B. M. Shawkat Ali, Amir H. Gandomi
Summary: Deep learning is revolutionizing evidence-based decision-making techniques and has the ability to overcome limitations posed by large datasets. However, as a multidisciplinary field that is still in its nascent phase, there is a limited number of articles that comprehensively review DL architectures. This paper aims to provide insights into state-of-the-art DL modelling techniques and their challenges and advantages.
ARTIFICIAL INTELLIGENCE REVIEW
(2023)
Article
Neurosciences
K. Martin-Chinea, J. Ortega, J. F. Gomez-Gonzalez, E. Pereda, J. Toledo, L. Acosta
Summary: This paper explores the use of LSTM and a low-cost wireless EEG device in real time to improve the BCI for individuals with impaired motor function. The results demonstrate higher classification accuracy with LSTM and an optimal time window of approximately 7 seconds for users to complete tasks.
COGNITIVE NEURODYNAMICS
(2023)
Article
Oncology
Aline Baltres, Zeina Al Masry, Ryad Zemouri, Severine Valmary-Degano, Laurent Arnould, Noureddine Zerhouni, Christine Devalland
Article
Computer Science, Artificial Intelligence
Antoine Proteau, Antoine Tahan, Ryad Zemouri, Marc Thomas
Summary: This article shows that a machine learning approach based on sensor data can be used to predict the quality measurement values of machined workpieces. A variational autoencoder regression model is proposed, and the Euclidean distance metric is correlated to the quality level in both the predicted and observed subsets.
JOURNAL OF INTELLIGENT MANUFACTURING
(2023)
Article
Pathology
Songhui Diao, Yinli Tian, Wanming Hu, Jiaxin Hou, Ricardo Lambo, Zhicheng Zhang, Yaoqin Xie, Xiu Nie, Fa Zhang, Daniel Racoceanu, Wenjian Qin
Summary: A weakly supervised framework based on a multiscale attention convolutional neural network (MSAN-CNN) was introduced for visual inspection of hepatocellular carcinoma cancer regions. The experimental results showed that this framework outperformed the single-scale detection method and had a very fast detection time.
AMERICAN JOURNAL OF PATHOLOGY
(2022)
Article
Medicine, General & Internal
Filippo Fraggetta, Vincenzo L'Imperio, David Ameisen, Rita Carvalho, Sabine Leh, Tim-Rasmus Kiehl, Mircea Serbanescu, Daniel Racoceanu, Vincenzo Della Mea, Antonio Polonia, Norman Zerbe, Catarina Eloy
Summary: The article discusses the implementation of digital pathology workflows and how to promote the adoption of digital pathology workflows in pathology laboratories. Consensus-based recommendations from experts cover every step of the digital workflow, emphasizing the importance of interoperability, automation, and process tracking.
Article
Energy & Fuels
Adaiton Oliveira-Filho, Ryad Zemouri, Philippe Cambron, Antoine Tahan
Summary: This article introduces a new condition-monitoring approach for wind turbines, which includes a built-in visualization tool to enhance interpretability of the model outcomes. The approach is based on a supervised implementation of the variational autoencoder model, enabling the projection of the wind turbine system onto a low-dimensional representation space. It provides a health indicator for early detection of abnormal conditions, a classifier for diagnosis status, and a visualization tool depicting the wind turbine condition as a trajectory in a 2D plot.
Article
Automation & Control Systems
Ryad Zemouri, Rony Ibrahim, Antoine Tahan
Summary: This study proposes a method using Variational Autoencoders (VAEs) and Sparse Dictionary Learning (SDL) for early anomaly detection of Hydraulic Turbine Generator Units (HTGU). Experimental tests show that this method can increase the sensitivity and robustness of anomaly detection.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Interdisciplinary Applications
Sagar Jose, Khanh T. P. Nguyen, Kamal Medjaher, Ryad Zemouri, Melanie Levesque, Antoine Tahan
Summary: Deep learning-based methods for industrial fault detection and diagnostics heavily rely on high-quality and sufficient quantity of condition monitoring data. However, in real-world industrial settings, data collection is often limited, resulting in insufficient data for training data-driven models. This work proposes a new approach that utilizes multimodal data and domain knowledge to address this issue. By leveraging complementary insights from different data modalities and domain expert knowledge, even when the data volume is low, the proposed methodology demonstrates the potential to enhance FDD accuracy and overcome sparse data challenges through a real industrial case study involving energy production systems.
COMPUTERS IN INDUSTRY
(2023)
Article
Engineering, Electrical & Electronic
Ryad Zemouri, Melanie Levesque
Summary: Condition monitoring of hydrogenerators is crucial for maintenance of power supply systems. Electrical insulation used in stator windings has been identified as a potent component for breakdown, with partial discharges (PDs) being one of the major causes. A intelligent system for automating PD source recognition is proposed, which utilizes a U-Net model to isolate the main sources of a phase-resolved PD signal (PRPD) and a set of deep learning models (DLMs) for automatic recognition of PD sources. Experimental results on real signals show promising accuracy for detecting different types of PDs. The proposed decision-making method provides a significant advantage over individual DLM performances.
IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION
(2023)
Proceedings Paper
Health Care Sciences & Services
Antoine Proteau, Ryad Zemouri, Antoine Tahan, Marc Thomas, Wafa Bounouara, Stephane Agnard
Summary: This paper presents a new dataset acquired in an industrial environment, which can be used to test and validate research work and accelerate technology transfer to industry.
2022 PROGNOSTICS AND HEALTH MANAGEMENT CONFERENCE, PHM-LONDON 2022
(2022)
Proceedings Paper
Health Care Sciences & Services
Ryad Zemouri, Melanie Levesque, Etienne Boucher, Mathieu Kirouac, Francois Lafleur, Simon Bernier, Arezki Merkhouf
Summary: This paper provides a synthesis of the latest research in the PHM domain using Variational Autoencoders (VAEs). It covers four main topics: data-driven soft sensors, fault detection, data resampling methods, and variational embedding. The paper reviews the theoretical foundations of VAEs, practical implementation tips, and showcases research done at Hydro-Quebec on hydro-generator failure detection using VAEs.
2022 PROGNOSTICS AND HEALTH MANAGEMENT CONFERENCE, PHM-LONDON 2022
(2022)
Proceedings Paper
Computer Science, Artificial Intelligence
Gabriel Jimenez, Anuradha Kar, Mehdi Ounissi, Lea Ingrassia, Susana Boluda, Benoit Delatour, Lev Stimmer, Daniel Racoceanu
Summary: This study introduces a DL-based method for semantic segmentation of tau lesions in brain tissues of AD patients, providing significant advantages for further stratification. Discussions on biomarkers, imaging modalities, and weak annotation challenges are crucial in this seminal research.
MEDICAL IMAGE COMPUTING AND COMPUTER ASSISTED INTERVENTION, MICCAI 2022, PT II
(2022)
Proceedings Paper
Computer Science, Information Systems
K. Manouskova, V Abadie, M. Ounissi, G. Jimenez, L. Stimmer, B. Delatour, S. Durrleman, D. Racoceanu
Summary: Tau proteins play a role in Alzheimer's disease, and detecting and segmenting the aggregates is crucial. This study presents a 5-step pipeline that improves state-of-the-art performances in detecting and segmenting tau protein aggregates, providing valuable insights in the field.
MEDICAL IMAGING 2022: DIGITAL AND COMPUTATIONAL PATHOLOGY
(2022)
Proceedings Paper
Engineering, Multidisciplinary
Olivier Kokoko, Claude Hudon, Melanie Levesque, Normand Amyot, Ryad Zemouri
Summary: This paper proposes a method for the automatic classification of PD patterns using CVAE and ten classifiers. By refining feature extraction rules, the accuracy level of classification can be significantly improved, raising questions about the performance of feature extraction rules and possibilities for better handling classification of large databases.
2021 IEEE INTERNATIONAL CONFERENCE ON PROGNOSTICS AND HEALTH MANAGEMENT (ICPHM)
(2021)
Proceedings Paper
Engineering, Electrical & Electronic
Claude Hudon, Melanie Levesque, Olivier Kokoko, Normand Amyot, Ryad Zemouri
Summary: Quantification of on-line Partial Discharge (PD) measurements is challenging due to instrumental characteristics, sensor types, sensor locations, and measurement procedures. Selecting the best parameter for quantification and differentiating PD sources are important issues, as simple quantification rules may not suffice. The study proposed using a Deep Convolutional Variational Autoencoder (DCVAE) to separate different types of PD sources, showing promising results in data analysis.
2021 IEEE ELECTRICAL INSULATION CONFERENCE (EIC)
(2021)
Article
Computer Science, Artificial Intelligence
Ryad Zemouri
MACHINE LEARNING AND KNOWLEDGE EXTRACTION
(2020)