Time-frequency methodologies in neurosciences

Abstract

This chapter presents a number of time-frequency (t,f) techniques that can provide advanced solutions to several problems in neuro-sciences with focus on the monitoring of brain abnormalities using EEG and other physiological modalities (t,f) characteristics as a diagnosis and prognosis tool. The methods presented illustrate the improved performance obtained by using a time-frequency approach to process EEG data, including a focus on detecting abnormalities in sick newborns in a Neonatal Intensive Care Unit (NICU) as well as mental health issues in elderlies.

The chapter starts by presenting methods for the assessment of Newborn EEG and ECG abnormalities using a time-frequency identification approach (Section 16.1). Next, the important question of (t,f) modeling of nonstationary signals is discussed with illustration on newborn EEGs (Section 16.2); Then, the use of (t,f) features for nonstationary signal classification is illustrated on an application to newborn EEG burst-suppression detection (Section 16.3); an application relevant to the elderly is described where a time-varying analysis of brain networks uses the EEG for the detection of Alzheimer disease (Section 16.4). Another method of time-frequency analysis is described that involves EEG noise reduction using the empirical mode decomposition(Section 16.5). Finally the chapter concludes with a discussion on other perspectives of using advanced (t,f) methods for medical diagnosis and prognosis in other areas of neurosciences (Section 16.6).