Abstract No. 720 Identification of irreversibly damaged brain tissue on computed tomography perfusion using convolutional neural network to assist selection for mechanical thrombectomy in ischemic stroke patients

Endovascular treatment of ischemic stroke has shown positive clinical outcomes. Further optimization requires identifying patients who will benefit from reperfusion. We propose using deep learning, specifically 3D convolutional neural networks (CNN), to identify infarcted tissue (core) on CT perfusion (CTP) with diffusion weighted imaging (DWI) MRI as gold standard for irreversible brain infarction and evaluate lesion size impact on the network ’s performance.

Source: Journal of Vascular and Interventional Radiology : JVIR - February 20, 2020 Category: Radiology Authors: R. Wang, K. Chang, H. Zhou, J. Wu, G. Cohan, M. Jayaraman, R. Huang, J. Boxerman, L. Yang, F. Hui, J. Woo, H. Bai Tags: Scientific Traditional Poster Source Type: research