A computational and visualisation tool for investigating associations between cardiac radiomics, risk factors and clinical data

Abstract

Identifying the correlations between radiomics and additional medical, health and lifestyle factors such as sex, age, BMI, etc. may help in discovering the significant hidden patterns of data and realizing the causes of the diseases. Also, knowing the risks in advance is a useful piece of supplementary information which may be used in addition to medical intervention resulting in preventative measures to reduce the level of risk or to control prescribed treatments. In the radiomics and the clinical outcomes datasets, it is hard to identify their correlations due to the complexity of data, computationally expensive and high number of possible combination among the choices. Therefore, data pre-processing to keep only the potential features and data cleaning to deal with missing or non-informative values are mandatory steps to perform. In addition, applying the powerful Machine Learning algorithms help to bring the results that even non-specialists in the field could discover and understand. This thesis facilitates the discovery of these correlations through the design and development of an intuitive and interactive web-based tool which dynamically displays the radiomic feature set alongside the additional medical, health and lifestyle factors feature set based on the contents of radiomics and clinical data files. The tool also provides a visualization of the correlation results in an easy to interpret and meaningful way allowing for effective exploration of any correlations in addition to cardiovascular risk score calculation.