AUTHOR=Teng Linyang , Ren Qianwei , Zhang Pingye , Wu Zhenzhou , Guo Wei , Ren Tianhua TITLE=Artificial Intelligence Can Effectively Predict Early Hematoma Expansion of Intracerebral Hemorrhage Analyzing Noncontrast Computed Tomography Image JOURNAL=Frontiers in Aging Neuroscience VOLUME=Volume 13 - 2021 YEAR=2021 URL=https://www.frontiersin.org/journals/aging-neuroscience/articles/10.3389/fnagi.2021.632138 DOI=10.3389/fnagi.2021.632138 ISSN=1663-4365 ABSTRACT=Significant early hematoma expansion in patients with intracerebral hemorrhage is an independent predictor of poor functional outcomes. Therefore, the identification of the possible enlargement of hematoma and prescription of treatment in accordance would facilitate the improvement of the prognosis of patients. Studies have demonstrated high sensitivity and specificity via computed tomographic angiography for predicting hematoma enlargement. However, the conditions of most of the patients do not allow complete examination in time. On the other hand, the use of non-contrast computed tomography is more convenient and popular in clinical settings. The signs of non-contrast computed tomography indicating hematoma enlargement needs to be recognized by doctors manually, as the outcome has high specificity, but low sensitivity. Therefore, it is extremely significant to predict hematoma enlargement more rapidly and accurately by using non-contrast computed tomography imaging. Considering these, we conducted a study to research whether hematoma enlargement can be predicted with non-contrast computed tomography images based on artificial intelligence. A total of 1899 non-contrast computed tomography images of cerebral hemorrhage patients were retrospectively analyzed to establish a predicting model, and 1117 to validate the model. And a total of 118 patients with intracerebral hemorrhage were selected based on the inclusion and exclusion criterions so as to validate the value of the model for clinical prediction. The baseline non-contrast computed tomography images within 6 h of intracerebral hemorrhage onset, and the second non-contrast computed tomography performed at 24 ± 3 h from the onset were used to evaluate the prediction of intracerebral hemorrhage growth. In the validation dataset 1, AUC was 0.778 (95% CI, 0.768-0.786), sensitivity was 0.818 (95% CI, 0.790-0.843), and specificity was 0.601 (95% CI, 0.565-0.632). In the validation dataset 2, AUC was 0.780 (95% CI, 0.761-0.798), sensitivity was 0.732 (95% CI, 0.682-0.788), and specificity was 0.709 (95% CI, 0.658-0.759). The sensitivity of intracerebral hemorrhage hematoma expansion as predicted by artificial intelligence imaging system was 89.3%, with the specificity of 77.8%, positive predictive value of 55.6%, negative predictive value of 95.9%, and a Yoden index of 0.671, which were much higher than those based on the manually labeled non-contrast computed tomography signs.