Artificial intelligence applications in stroke

Abstract

Machine learning and other artificial intelligence techniques have been extensively used in a large number of medical applications, in diagnosis, treatment selection, mining of electronic health records, genetics, and image processing, among several others. Machine learning techniques can be used to infer predictive models, from labeled data, in many areas of medicine, including stroke. In particular, in patients who suffered from stroke, machine learning can be used to perform or improve outcome prediction, lesion segmentation, and treatment assessment, among others. Machine learning algorithms can be based on a number of different approaches. In this chapter, we cover the symbolic, statistical, similarity-based, and connectionist approaches, which have different properties and trade-offs. Recently, a set of techniques generally known as deep learning have increased significantly the range of applicability of machine learning methods, which are now able to deal with several problems characterized by high-dimensional data, such as images and videos, outperforming even experts in tasks such as lesion segmentation and outcome prediction. We provide an overview of the different machine learning methods used in the stroke area and some of the applications proposed by researchers in the field.