AUTHOR=Yi Yan , Mao Li , Wang Cheng , Guo Yubo , Luo Xiao , Jia Donggang , Lei Yi , Pan Judong , Li Jiayue , Li Shufang , Li Xiu-Li , Jin Zhengyu , Wang Yining TITLE=Advanced Warning of Aortic Dissection on Non-Contrast CT: The Combination of Deep Learning and Morphological Characteristics JOURNAL=Frontiers in Cardiovascular Medicine VOLUME=Volume 8 - 2021 YEAR=2022 URL=https://www.frontiersin.org/journals/cardiovascular-medicine/articles/10.3389/fcvm.2021.762958 DOI=10.3389/fcvm.2021.762958 ISSN=2297-055X ABSTRACT=Background:Identification of the Aortic dissection (AD) at baseline plays a crucial role in clinical practice. Non-contrast CT scans are widely available, convenient, and easy to perform. However, the detection of AD on non-contrast CT scans by radiologists is lack of sensitivity and suboptimal at present. Methods: A total of 452 patients underwent aortic CT angiography (CTA) were enrolled retrospectively from two medical centers in China to form an internal cohort (341 patients, 139 AD, 202 non-AD) and an external testing cohort (111 patients, 46 AD, 65 non-AD). The internal cohort was divided into the training cohort (n=238), validation cohort (n=35) and internal testing cohort (n=68). The morphologic characteristics were extracted from the aortic segmentation. A Deep-Integrated model that based on the Gaussian Naive Bayes algorithm were built to differentiate AD from non-AD, using the combination of the 3D deep-learning model score and the morphologic characteristics. The areas under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity were used to evaluate the model performance. The proposed model was also compared with the subjective assessment of radiologists. Results: After the combination of all morphologic characteristics, our proposed Deep-Integrated model significantly outperformed the 3D deep-learning model (AUC: 0. 948 vs. 0.803 on the internal testing cohort, 0.969 vs. 0.814 on the external testing cohort, both p<0.05). The accuracy, sensitivity, and specificity of our model reached 0.897, 0.862, and 0.923 on the internal testing cohort, and 0.730, 0.978, and 0.554 on the external testing cohort, respectively. The accuracy for AD detection showed no significant difference between our model and the radiologists (all p > 0.05). Conclusions: The proposed model presented good performance on the AD detection on non-contrast CT scans, thus early diagnosis as well as prompt treatment would be available.