This postdoctoral research aims to establish new clinical practices for PE management by learning innovative biomarkers, in close partnership with leading clinicians. Our primary goal is to leverage state-of-the-art multi-modal contrastive learning on our unique dataset of over 400 patients to develop an embedding unifying CTPA, tabular patient data and the aforementioned biomarkers.
This embedding will first be used to develop novel risk stratification models. Then, the learned embedding will be analyzed to evaluate the individual contribution of each biomarker towards the stratification task. This will open the way to larger-scale clinical studies of these biomarkers to define future PE clinical practices.
More info on the attached pdf.