Abstract #1792

Accuracy of quantified 23Na MRI in ischemic Stroke with varying undersampling Factors and CNN Postprocessing

Anne Adlung¹, Nadia Karina Paschke¹, Alena-Kathrin Golla^1,2, Dominik Bauer^1,2, Sherif Mohamed³, Melina Samartzi⁴, Marc Fatar⁴, Eva Neumaier Probst³, Frank Gerrit Zöllner^1,2, and Lothar Rudi Schad¹

¹Computer Assisted Clinical Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, ²Mannheim Institute for Intelligent System in Medicine, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, ³Department of Neuroradiology, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany, ⁴Department of Neurology, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany

This study investigates factors of k-space undersampling for which CNN postprocessing is able to improve ²³Na MRI data. Data from 53 patients with ischemic stroke was included and image reconstruction was performed with full k-space data (FI) and with k-space data that was reduced (RI) by different factors (S = 2, 4, 5 and 10). Postprocessing with a convolutional neural network was applied to the highly undersampled ²³Na MRI data. The CNN was able to significantly improve SNR and SSIM for all S with both loss functions. CNN postprocessing could enable significant reduction of ²³Na MRI data acquisition time.

This abstract and the presentation materials are available to members only; a login is required.

Join Here