Photo Credit: Tippapatt
The following is a summary of “Prediction of Tissue Outcome in Acute Ischemic Stroke based on Single-Phase CT Angiography at Admission,” published in the March 2024 issue of Neurology by Palsson et al.
Accurate prediction of tissue recovery following reperfusion in acute ischemic stroke helps identify patients who could benefit most from mechanical thrombectomy (MT).
Researchers started a retrospective study to develop a deep learning model for predicting follow-up infarct location and extent using only widely available acute single-phase computed tomography angiography (CTA) data, eliminating the need for CT perfusion (CTP) imaging.
They screened all patients with acute large vessel occlusion of the anterior circulation treated (December 2015 and December 2020) (N=404) and included 238 patients undergoing MT with successful reperfusion for final analysis. Ground truth infarct lesions were segmented on 24-hour follow-up CT scans. Pre-processed CTA images were input for a U-Net-based Convolutional Neural Network (CNN) trained for lesion prediction, enhanced with a spatial and channel-wise squeeze-and-excitation block. Post-processing removed small predicted lesion components. The model was evaluated using a 5-fold cross-validation and a separate test set with the Dice Similarity Coefficient (DSC) as the primary metric and average volume error as the secondary metric.
The results showed that the mean ± standard deviation of the test set DSC overall folds after post-processing was 0.35±0.2, and the mean test set average volume error was 11.5 milliliters. The performance demonstrated uniformity across models, with the highest-scoring model achieving a DSC of 0.37±0.2 after post-processing and the model with the lowest average volume error yielding 3.9 milliliters.
They concluded that acute CTA imaging alone predicted 24-hour infarcts, similar to CT perfusion, potentially making AI-based prediction more feasible.
Source: frontiersin.org/journals/neurology/articles/10.3389/fneur.2024.1330497/abstract
