Combining Investigation of Imaging Markers (Island Sign and Blend Sign) and Clinical Factors in Predicting Hematoma Expansion of Intracerebral Hemorrhage in the Basal Ganglia

BACKGROUND AND OBJECTIVE: Intracerebral hemorrhage (ICH) is the most difficult-to-treat form of stroke and accounts for about 10%-30% of all strokes worldwide. Hematoma expansion (HE), which occurs in one third of patients with ICH, is strongly predictive of worse prognosis and potentially preventable if high-risk patients are identified in the early phase of ICH. We summarized data from recent studies on HE prediction and classified those potential indicators into 2 categories: 1) clinical and laboratory and 2) radiographic. Therefore, we aimed to identify the accuracy of 1, that is, the value of combining predictors in predicting HE of ICH in basal ganglia. METHODS: We retrospectively investigated the clinical database of Qinghai Provincial People's Hospital for patients with ICH aged >18 years between January 2015 and January 2018. As inclusion criteria, we defined 1) ICH diagnosed on noncontrast computed tomography (CT); 2) noncontrast CT performed on enrollment within 6 hours after onset of symptoms; 3) follow-up CT scan performed within 24 hours after the baseline CT scan; and 4) all of the primary hematoma was located in the basal ganglia. Univariate and multivariate logistic regression analysis were used to analyze the potential HE predictors, and then receiver operating characteristic curves were used to evaluate the L (the value of combining predictors) of imaging markers and clinical factors in predicting HE. RESULTS: Of the 99 patients with HE, island sign was present in 48.48% (48/99) of patients and blend sign was present in 34.34% (34/99) of patients. Multivariate logistic regression analysis identified time to baseline CT scan (odds ratio [OR] 1.574; 95% confidence interval [CI] 1.205-2.054; P = 0.001), baseline hematoma volume (P = 0.001), presence of island sign (OR 11.247; 95% CI 4.701-26.909; P = 0.000), presence of blend sign (OR 3.104; 95% CI 1.425-6.765; P = 0.004), anticoagulants use or international normalized ratio >1.5 (OR 2.755; 95% CI 1.072-7.082; P = 0.035), and intraventricular hemorrhage (OR 2.351; 95% CI 1.066-5.187; P = 0.034) as independent predictors of HE. The sensitivity and specificity of L (value of combining predictors) were 88.89% and 80.84%, respectively; the area under the curve was 0.918. CONCLUSIONS: The findings indicated that the ability of L to predict HE was much more excellent than these 6 predictors alone. L showed a high association with HE, with an accuracy of 91.8%, and was a reliable value of combining predictors in terms of predicting HE. L may serve as a promising, noninvasive tool for clinical therapeutic strategy.