Tuesday 17 December 2019 16:00 – 17:00 A. Payatakes Seminar Room
“Application of pattern recognition and texture analysis methods on cancer imaging data”
Dr Maria Venianaki Institute of Computer Science (ICS)
Abstract
Cancer research has significantly advanced in recent years mainly through developments in medical genomics and bioinformatics. From the imaging perspective, imaging biomarkers have been proposed in numerous studies as a cost-effective, non-invasive method for cancer diagnosis, monitoring and therapy outcome prediction, as they can provide important anatomical and functional information at an early stage of therapy. However, the extraction of imaging biomarkers is still an open problem, as there are no standard imaging protocols or established methods. Pharmacokinetic models are the most commonly used technique for analyzing dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data and extracting imaging biomarkers. However, these models have several limitations such as their computational complexity, which often results in high variability of the computed biomarkers.
In this talk, we will first present the fundamentals of the DCE-MRI technique, which is a noninvasive imaging method that can provide information concerning tissue oxygenation and vascularization at high spatial resolution. We will then discuss data-driven, model-free biomarker extraction strategies relying on the classification of time intensity curves. Specifically, pattern recognition techniques for the extraction of enhancement patterns from DCE-MRI data will be presented. Results from different cancer imaging datasets will be shown as different case studies. Finally, we will discuss the extraction of texture-based imaging biomarkers from DCEMRI data and their application on the prediction of treatment response.