Big Data, Small Vessels

Abstract

In recent years, the transformative potential of data science in reshaping various sectors, including healthcare, has become evident. This thesis explores the application of data science in cardiovascular research, which is heavily impacted by the prevalence and complexity of cardiovascular diseases (CVD), the leading global cause of death. The research addresses both large vessel disease (LVD) and small vessel disease (SVD), which are influenced by a combination of genetic, environmental, and lifestyle factors. Personalized treatment approaches are crucial due to the multifaceted nature of these diseases.

The core focus of the thesis is the utilization of "big data", specifically electronic health records, raw data from routine hematology analyzers, and large protein panels, which hold potential value for cardiovascular research. Advanced data science techniques, such as machine learning, natural language processing (NLP), and network analysis, are employed to analyze high-dimensional data, handle missing information, extract insights from unstructured textual data, and unravel biological networks.

Chapter Two describes the reuse of routine care data, especially laboratory data, for research purposes. It delves into the challenges and solutions related to data quality and availability, emphasizing the crucial role of multidisciplinary human expertise in data interpretation.

Chapter Three investigates the risks of acute cardiovascular diseases in patients post-carotid endarterectomy. It highlights the potential role of specific blood cell characteristics, such as neutrophil size, in inflammation and disease risk.

Chapter Four examines the sex-specific biological processes related to the success of endovascular thrombectomy in treating acute ischemic strokes. It uncovers significant differences in blood cell characteristics between males and females.

Chapter Five reveals a potential underreporting of cognitive problems in heart failure patients, identified through text mining of clinical anamneses from routine care.

Chapter Six details the development of a sex-specific decision-support algorithm trained on routine care data, to enhance the diagnostic process and cost-effectiveness by ruling out coronary artery disease before cardiac imaging in patients with chest discomfort.

Chapter Seven validates these algorithms across various settings and populations, confirming their effectiveness in accurately ruling out artery disease before cardiac imaging.

Chapter Eight utilizes cluster analysis to decipher the biological mechanisms of cerebral small vessel disease, identifying a protein network associated with blood clotting and inflammation.

Chapter Nine investigates protein networks linked to reduced cerebral blood flow and future cardiovascular events in patients with cardiovascular disease, identifying proteins related to extracellular matrix organization and inflammation.

Chapter Ten highlights the potential of data science in cardiovascular research. It discusses challenges in data cleaning, the necessity of robust IT infrastructure and skilled personnel, and the importance of interdisciplinary collaboration. The chapter advocates the Dataïsm paradigm, which considers data as central to understanding human health and well-being, and promotes a patient-centric approach where algorithms support, but do not replace, clinical judgment. Additionally, the chapter explores the future potential of data science in cardiovascular medicine, including the use of speech recognition systems and real-time data collection from wearable devices.