Читать книгу EEG Signal Processing and Machine Learning - Saeid Sanei - Страница 9

Preface to the Second Edition

Brain research has reached a considerable level of maturity due, for example, to having access to: a wealth of recording and screening resources; availability of substantial data banks; advanced data processing algorithms; and emerging artificial intelligence (AI) for making more accurate clinical diagnosis. Neurotechnology is also now being exploited to design revolutionary interfaces to guide artificial prostheses for human rehabilitation. Moreover, the technology for brain repair, communications between live and AI‐based body parts, mind reading, and intelligent recordings together with the use of virtual and augmented reality domains is advancing remarkably. The advances in brain research will soon make the Internet‐of‐brains feasible and enable fully monitoring the body for personal medicine purposes.

To progress this fast‐growing technology, the demand for electroencephalography (EEG) data, as a widely accessible, informative, flexible, and expandable brain screening modality, together with suitable approaches in EEG processing, is rising dramatically.

Automatic clinical diagnosis requires signal processing and machine learning algorithms to bring more insight into interpretation of the data, devising a treatment plan, and defining the path for achieving personalized medicine which is the goal of future healthcare systems. EEG is of particular interest to researchers due to its very rich information content and its relation to the entire body function.

EEG signals represent three fundamental activities in the brain: firstly, they show the normal brain rhythms which exist in the EEGs of healthy subjects and indicate the human states such as awake and sleep; secondly, they demonstrate the brain responses to audio, visual, and somatosensory excitations, whose variations can represent the brain performance in the cases of mental fatigue, learning, and memory load; and thirdly, the communications between various brain zones which can change due to ageing, dementia, and many other factors. The study of these three aspects of EEG is the focus of this book.

Most of the concepts in single channel or multichannel EEG signal processing have their origin in distinct application areas such as communication, seismic, speech and music signal processing. EEG signals are generally slow‐varying waveforms and therefore, similar to many other physiological signals, can be processed online without much computational effort.

This second edition of the book EEG Signal Processing, first published in 2007, highlights the major impact machine learning is now having on EEG analysis. This has been made possible by the recent developments in data analysis: firstly, due to the availability of supercomputers, powerful graphic cards, large volume computer clusters, and memory space within the public cloud, and secondly, due to introducing powerful classification algorithms such as deep neural networks (DNNs) which are suitable for numerous applications in brain–computer interfacing, mental task evaluation, brain disorder/disease recognition, and many others.

This edition is inclusive and comprehensive, encompassing almost all methodologies in EEG processing and learning together with their diverse applications. It is not only the result of the endeavours of our research teams, but also an encyclopaedia of the most recent works in EEG signal processing, machine learning, and their applications. Hence, this edition covers a wider, deeper, and richer content thereby alleviating the shortcomings in the first edition of this book. As such, this edition can be used as a reference by researchers in bioengineering, neuroscience, psychiatry, neuroimaging, and brain–computer interfacing. It can also be used for teaching bioengineering and neuroengineering at different university levels.

In this second edition, the number of chapters has increased from 7 to 18 by covering and extending the content in each chapter and adding many new topics for analysis of EEG signals including: (i) offering deeper understanding and insight into the generation of EEG signals and modelling the brain EEG generators in Chapters 2 and 3, (ii) being more inclusive in the domains of theoretical and practical aspects in EEG single‐ and multichannel signal processing including static and dynamic systems within multimodal and multiway mathematical models in Chapters 3–6, and (iii) providing a comprehensive and detailed approach to AI, particularly machine learning approaches, starting from traditional crisp classification to advanced deep feature learning approaches in Chapter 7.

Chapter 8 addresses brain coherency, synchrony, and connectivity. This chapter introduces a completely new topic of cooperative learning and adaptive filtering into the domain of brain connectivity and its applications. Chapter 9 introduces the brain response to audio, visual, and tactile events when they are regularly presented or targeted in an odd ball paradigm. The important topic of brain source localization, using both forward and inverse problems, is addressed in Chapter 10. A vast range of applications of these three chapters is given in the ensuing chapters.

From Chapter 11 onwards, more practical and clinically demanding approaches are discussed with the help and direct application of the theoretical developments in the previous chapters. Seizure and epileptic waveforms are studied comprehensively in Chapter 11. This chapter includes a very innovative approach using DNNs to model the pathways between the generators of epileptiform discharges to the scalp electrode recordings.

The fundamental objectives of new and advanced materials included in Chapters 12–14 are to assess the cortical brain waves, the coherency and connectivity within various brain zones, and the brain responses to different stimuli while the subject is in different states of awake, sleep, mentally tired, and under different emotions. Chapters 15 and 16 introduce the state‐of the‐art techniques in signal processing and machine learning for recognition of degenerative diseases and neurodevelopmental disorders respectively.

Brain–computer interfacing benefiting from a wealth of new neurotechnology together with applications of advanced AI systems for rehabilitation, computer gaming, and eventually brain communications purposes is comprehensively covered in Chapter 17 which concludes application of EEG signals and systems. Finally, Chapter 18 shows how EEG can be combined with other simultaneously recorded functional neuroimaging data. It introduces a number of applications where EEG can be combined with functional magnetic resonance images (fMRI) and functional near‐infrared spectroscopy (fNIRS) images, to exploit their high spatial resolutions for enhancing the overall diagnostic performance.

In the treatment of various topics covered within this research monograph it is assumed that the reader has a background in the fundamentals of digital signal processing and machine learning and wishes to focus on EEG analysis. It is hoped that the concepts covered in each chapter provide a solid foundation for future research and development in the field.

As we concluded in the first edition, we do wish to stress that in this book there is no attempt to challenge previous clinical or diagnostic knowledge. Instead, the tools and algorithms described in this book can, we believe, potentially enhance the significant information within EEG signals and thereby aid physicians and ultimately provide more cost effective and efficient diagnostic tools.

Both authors wish to thank most sincerely our Research Associates and PhD students who have contributed so much to the materials in this work.

Saeid Sanei and Jonathon A. Chambers