Artificial intelligence in neurology

Abstract

Artificial intelligence (AI) can process and interpret large amounts of data and thus shows great potential in neurology. Many neurological diseases require extensive multimodal diagnostics and personalized therapy that can be automated or optimized by AI. AI comprises complex algorithms. This article aims to provide an overview of the basic terminology, algorithms, and applications of AI in neurology. There is an increasingly large number of different AI models. Currently, neural networks and transformers are particularly powerful. Neural networks can, for example, provide diagnostic or prognostic assessments by processing static input data such as radiological images. They use many small computational units, neurons, arranged into networks. Transformers can process and output sequential data such as text. They can thus, for example, provide tentative diagnoses or formulate reports of findings based on anamnesis interviews. Transformers calculate the relationship of the individual text segments to each other so that these can be considered during processing. A basic understanding of the applications and functionality of AI, but also of the challenges and limitations, is crucial for its successful use in neurological research and practice.