Prediction of tissue outcome in acute ischemic stroke based on single-phase CT angiography at admission

• DOI: 10.3389/fneur.2024.1330497
• Publication Date: 2024-03-19T04:51:39Z
• Authors: Frosti Palsson; Nils D. Forkert; Lukas Meyer; Gabriel Broocks; Fabian Flottmann; Máté E. Maros; Matthias Bechstein; Laurens Winkelmeier; Eckhard Schlemm; Jens Fiehler; Susanne Gellißen; Helge C. Kniep
• Keywords: Penumbra; Stroke; Acute stroke

Introduction In acute ischemic stroke, prediction of the tissue outcome after reperfusion can be used to identify patients that might benefit from mechanical thrombectomy (MT). The aim this work was develop a deep learning model predict follow-up infarct location and extent exclusively based on single-phase computed tomography angiography (CTA) datasets. comparison CT perfusion (CTP), CTA imaging is more widely available, less prone artifacts, established standard care in stroke protocols. Furthermore, recent RCTs have shown also with large infarctions MT, which not been selected for MT CTP core/penumbra mismatch analysis. Methods All vessel occlusion anterior circulation treated at our institution between 12/2015 12/2020 were screened ( N = 404) 238 undergoing successful included final Ground truth lesions segmented 24 h scans. Pre-processed images as input U-Net-based convolutional neural network trained lesion prediction, enhanced spatial channel-wise squeeze-and-excitation block. Post-processing applied remove small predicted components. evaluated using 5-fold cross-validation separate test set Dice similarity coefficient (DSC) primary metric average volume error secondary metric. Results mean ± deviation DSC over all folds post-processing 0.35 0.2 11.5 mL. performance relatively uniform across models best according achieved score 0.37 terms yielded 3.9 Conclusion feasible measures comparable results CTP-based algorithms reported other studies. proposed method pave way wider acceptance, feasibility, applicability artificial intelligence.