AI-based EEG analysis: new technology and the path to clinical adoption

This review article introduces the challenges and potential benefits of the use of Artificial Intelligence (AI) and Machine Learning (ML) models in the analysis of the human electroencephalogram (EEG) in clinical practice. The integration of AI-based tools into clinical workflow has the potential to offer new insights, increase efficiency and essentially improve outcomes, but care must be taken during the development, validation, integration and interpretation of such tools to ensure their viability and efficacy, lest we risk introducing complexity and potentially harmful misinterpretations. The primary objective of this review is to introduce the key concepts of AI-based clinical tools to a non-technical audience, and to present the criteria for successful integration of such tools. Drawing from the authors' collective 35-year experience meeting medical device regulatory requirements for new technology, as well as a broad sampling of the current literature, this review offers insights into the challenges and opportunities of AI-based tools in clinical practice. The analysis is generalized to AI-based software, but key examples of EEG measurement and usage in clinical neurology in the treatment of epilepsy and other disorders are presented. The authors suggest that while digital technology has the potential to revolutionize clinical practice, specifically in the monitoring of epileptic patients, it also raises issues related to digital inequality, automation bias and basic performance. The implications of this review point to the need for more transparent development and clinical integration, based fundamentally on the collaboration between technology innovators and clinical practitioners, to allow safe and effective adoption of new technologies with the maximum likelihood of clinical success.