With an estimated global incidence of 11 million patients per year, research involving ischemic stroke requires the collection and analysis of massive data sets affected by innumerable variables. Landmark studies that have historically shaped the foundation of our understanding of ischemic stroke and the development of management protocols have been derived from only a miniscule fraction of a percent of the entire population due to feasibility and capability. Machine learning provides an opportunity to capture data from an extraordinarily larger cohort size, which can be applied to training models to formulate algorithms to forecast outcomes with unparalleled accuracy and efficiency. The paradigm-shifting integration of machine learning in other industries, i.e. robotics, finance, and marketing, foreshadows its inevitable application to large population-based clinical research and practice. While prior multi-center studies have relied heavily on catalogued datasets requiring substantial manpower, the recent development of modern statistical methods can potentially expand the available quantity and quality of clinical data. In conjunction with data mining, machine learning has allowed automated extraction of clinical information from imaging, surgical videos, and electronic medical records to identify previously unseen patterns and create prediction models. Recently, it’s use in real-time detection of large vessel occlusion has streamlined health care delivery to a level of efficiency previously unmatched. The application of machine learning in ischemic stroke research – data acquisition, image evaluation, and prediction models – has the potential to reduce human error and increase reproducibility, accuracy, and precision with an unprecedented degree of power. However, one of the challenges with this integration remains the methods in which machine learning is utilized. Given the novelty of machine learning in clinical research, there remains significant variations in the application of machine learning tools and algorithms. The focus of the research topic is to provide a platform to compare the merits of various learning approaches – supervised, semi-supervised, unsupervised, self-learning – and the performances of various models.

Machine learning (ML) and deep learning (DL) have become essential tools in healthcare. They are capable of processing enormous amounts of data to find patterns and are also adopted into methods that manage and make sense of healthcare data, either electronic healthcare records or medical imagery. This book explores how ML/DL can assist neurologists in identifying, classifying or predicting neurological problems that require neuroimaging. With the ability to model high-dimensional datasets, supervised learning algorithms can help in relating brain images to behavioral or clinical observations and unsupervised learning can uncover hidden structures/patterns in images. Bringing together artificial intelligence (AI) experts as well as medical practitioners, these chapters cover the majority of neuro problems that use neuroimaging for diagnosis, along with case studies and directions for future research.

Omics-based approaches have emerged as powerful tools in stroke research, revolutionizing our understanding of the underlying molecular mechanisms and potential therapeutic targets. These approaches encompass various disciplines such as genomics, transcriptomics, proteomics, metabolomics, radiomics, and epigenomics, enabling comprehensive analysis of biological and imaging markers and their interactions. Through genomics, researchers can identify genetic variants associated with stroke susceptibility, offering insights into individual risk factors and personalized medicine. Transcriptomics allows the investigation of gene expression patterns, highlighting key molecular pathways involved in stroke pathology and providing potential targets for intervention. Proteomics aids in the identification and quantification of proteins associated with stroke, aiding in the discovery of novel biomarkers and therapeutic targets. Metabolomics explores the metabolites involved in stroke pathophysiology, shedding light on metabolic alterations and potential therapeutic strategies. Radiomics involves the extraction and analysis of a multitude of quantitative features from medical imaging data, such as CT or MRI scans serving as potential imaging biomarkers, contributing to risk stratification and the identification of novel insights into stroke pathophysiology. Finally, epigenomics investigates modifications in gene expression without changing the DNA sequence, uncovering epigenetic mechanisms underlying stroke susceptibility and recovery. By integrating and analyzing data from these omics platforms, researchers can gain a comprehensive understanding of stroke pathogenesis, paving the way for the development of innovative diagnostic tools and effective therapeutic interventions.

This book constitutes the refereed proceedings of the 24th Conference on Medical Image Understanding and Analysis, MIUA 2020, held in July 2020. Due to COVID-19 pandemic the conference was held virtually. The 29 full papers and 5 short papers presented were carefully reviewed and selected from 70 submissions. They were organized according to following topical sections: ​image segmentation; image registration, reconstruction and enhancement; radiomics, predictive models, and quantitative imaging biomarkers; ocular imaging analysis; biomedical simulation and modelling.