Multimodal EEG–MRI signal processing approaches for stroke and rehabilitation: a narrative review

Abstract

The occurrence of stroke continues to be one of the leading causes of disability in the world today, stressing the importance of a correct diagnosis and rehabilitation strategy that takes into account the dynamic process of brain reorganization that follows such an event. While electroencephalography (EEG) is used to study neuronal activity on the millisecond scale, magnetic resonance imaging (MRI) provides high-resolution structural and functional images. While it remains that the choice for the acute setting as an imaging modality is MRI/CT to inform treatment decisions, it can increasingly be argued that linking MRI to EEG becomes more valuable for prognosis, intensity assessment, and rehabilitation at this level of care. In this narrative review, we discuss recent advancements in EEG-MRI signal processing for ischemic and hemorrhagic stroke, with particular attention to preprocessing, artifact removal, feature extraction, multimodal data fusion, and machine learning methods. In addition to presenting the existing methods for stroke-related multimodal applications, we highlight current challenges for translating these techniques from research to clinical practice, which include the limited availability of publicly available multimodal data, a lack of longitudinal studies covering acute to chronic phases of recovery, inconsistency in standardization of acquisition and preprocessing protocols, and neglecting applications related to rehabilitation. It is important to note that this review provides an overview of how emerging technologies can address the problems mentioned above. This review highlights developments in neuroimaging, signal processing, artificial intelligence, and rehabilitation science, and discusses the potential applications of these techniques in stroke treatment.