Cardiac disease is the leading cause of death worldwide, and according to the World Health Organization (WHO) more than 500 million people worldwide live with cardiovascular disease. Patients are often examined using echocardiography (ultrasound imaging of the heart) because this imaging technique is radiation-free, non-invasive, affordable, and can provide both structural and functional information about the heart. However, the technique is subject to challenges related to image acquisition and quality, time-demanding analysis, and large variations in the results of analysis. The increased burden of cardiovascular disease on the healthcare system causes demands for more efficient and precise analysis methods that can provide improved diagnostic and prognostic support for clinicians.
In this PhD project, we aim to investigate whether analysis methods based on machine learning (ML) can yield results as good as the reference methods for the diagnosis and prognosis of various heart diseases. Furthermore, we will attempt to develop methods to combine information from ultrasound images with other sources of information about the patient’s condition, approaches known as holistic or multimodal artificial intelligence (AI). Information such as that from patient characteristics, electrocardiography (ECG), and blood pressure measurements can be valuable and complementary to the ultrasound images. Throughout the project, we will aim to use techniques for explainability (eXplainable AI, XAI) to increase the transparency of the methods and help users understand how the AI models are processing information and producing predictions.