LEARNING-BASED DETECTION OF FLOW DIVERTERS IN CEREBRAL IMAGES

We propose a machine learning-based method to automatically detect flow diverters in cerebral C-arm CT images. An appearance detector is learned to generate hypotheses of a flow diverter's location in a volumetric image. A probabilistic framework incorporating a local appearance and shape model is developed to trace the flow diverter. Promising results have been obtained on clinical data. The proposed method provides a potential solution to the automation of cerebral aneurysm treatment workflow and in particular the post-operative assessment of flow diverter placement.