Combining radiomics and disease state index for interactive patient space visualization

Abstract

With the increasing number of data routinely acquired in clinical practice, a great variety of predictive models for automated medical diagnosis based on quantitative feature extraction are being implemented. Domain analysts should be able to interact and explore through different views to make further discovery of, and insights into, the quantitative data. They can obtain a better understanding of the data and their structures and contribute their domain expertise to the knowledge discovery process. For this research, we will take advantage of the use of radiomics for the acquisition of large amounts of relevant data from cardiac images that typically fail to be appreciated by the naked eye. The combination with Disease State Index, a predictive model that involves not only a diagnosis but a sense of progression in the disease, leads to a 3D space where the patients are explored and compared, as well as their individual features. This configuration will result in a clinical tool that allows the user to explore radiomics features in a serie of interactive panels and take supervised decisions, but also is useful for automatic diagnosis and patient stratification. Last but not least, it draws conclusions about the relevance of different radiomics classes, segmentations, and cardiac phases for automatic heart disease classification.