Application of self-designed collateral circulation quantitative score based on multi-task learning vascular segmentation in sCTA assessment of collateral circulation in acute ischemic stroke / 西安交通大学学报(医学版)

ABSTRACT

【Objective】 To evaluate the clinical use of the baseline CT angiography (CTA) quantitative score (self-designed collateral circulation quantitative, SD-CCQ) in determining the collateral circulation compensation status in patients with acute ischemic stroke (AIS), as well as the reliability and accuracy of the SD-CCQ score and the Alberta Stroke Program Early CT Score (ASPECTS). 【Methods】 Retrospective analysis was made on the clinical and imaging data, including CT, CTA and DWI image data, of 84 patients who were admitted for acute ischemic stroke to the Department of Neurorehabilitation of Zhongshan Hospital of Traditional Chinese Medicine from January 2020 to December 2022.Their CTA source images were annotated using a multi-task deep learning method for vascular segmentation. The ASPECTS score and SD-CCQ score were then applied to the CTA images following vascular segmentation in order to assess the collateral circulation compensation of AIS patients. The Kappa test was used to assess the consistency of the two methods used to assess collateral circulation, and the multifactorial Logistic regression analysis was used to examine the relationship between the SD-CCQ and the prognosis of the AIS patients. 【Results】 ASPECTS score had good consistency with SD-CCQ score in evaluating collateral circulation in AIS patients (κ=0.65, P<0.001), and the diagnostic accuracy of the latter for benign collateral circulation in AIS was 96.15%. Logistic regression analysis showed that the new collateral circulation score, baseline NIHSS, and DWI infarct volume were the main factors affecting the long-term prognosis of AIS patients. 【Conclusion】 The new scoring system SD-CCQ can be used to evaluate the compensatory status of collateral circulation in AIS patients, which may help in clinical treatment decision-making and prognosis prediction.