|Recent studies show that patient outcome following ischemic stroke can be improved considerably with proper treatment selection in the acute phase. In the setting of acute stroke, CTA imaging data provide important information that can be used for this treatment selection. However, this information is currently sub-optimally used as the imaging data are only assessed qualitatively. Accurate and objective quantification of CTA imaging biomarkers, combined with advanced decision support models has the potential to greatly improve current clinical practice. The aim of the project is therefore to develop and validate methods for the automated analysis of CTA data acquired in the
acute stroke setting and to determine the added value of this analysis for supporting treatment selection for patients with acute ischemic stroke.
In the first phase of the project, deep learning based models have been developed for automated image analysis. These generic models can be used for extraction of quantitative imaging biomarkers, as well as for prediction purposes. First results of the training of these models demonstrate that these models indeed can be used to extract quantitative imaging biomarkers such as collateral scores.
In the second phase of the project, we intend to extend the amount of training data, to further improve the training results for biomarker extraction and to perform a validation study. In addition, we intend to further develop and train the same models for prediction of treatment outcome.