Transcranial Ultrasonography in the Detection of Cerebrovascular Accident (CVA): A Systematic Review and Bivariate Random-Effects Meta-Analysis

Abstract

Abstract: Background/Objectives: Stroke remains a leading cause of disability and death worldwide, with rapid diagnosis critical for effective treatment. Transcranial ultrasonography offers a potentially valuable diagnostic tool, particularly in resource-limited settings. This study aimed to evaluate the diagnostic accuracy of transcranial Doppler (TCD) and transcranial color-coded duplex sonography (TCCS) for detecting ischemic and haemorrhagic stroke. Methods: We conducted a systematic review and bivariate random-effects meta-analysis following PRISMA guidelines (PROSPERO CRD42023471425). Three databases (Scopus, PubMed, Web of Science) were searched from inception to November 20, 2022. We included studies assessing adult (≥18 years) stroke patients using TCD or TCCS compared to reference standards (MRI, CT, or angiography). Exclusion criteria included paediatric populations, animal studies, vasospasm assessments, and studies with <10 participants. Four reviewers independently screened studies, extracted data, and assessed risk of bias using NIH tools. Primary outcomes were pooled sensitivity and specificity analysed using bivariate random-effects models; this technique was chosen as it jointly analyses sensitivity and specificity using a random-effects model that accounts for their correlation and between-study variability. Results: From 16,397 records, 26 studies met inclusion criteria, with 13 studies (n=802 ultrasound examinations) included in meta-analysis. Overall diagnostic accuracy showed sensitivity of 81.1% (95% CI 73.1-87.2%) and specificity of 85.5% (95% CI 71.4-93.3%), with an AUC of 0.874. TCCS demonstrated higher sensitivity for haemorrhagic stroke (87.8%) than for ischemic stroke (77.2%). TCCS showed sensitivity of 74.8% (95% CI 66.4-81.6%) and specificity of 85.2% (95% CI 63.0-95.1%), with an AUC of 0.796. Risk of bias was low in 30.6% of studies, moderate in 50%, and high in 19.4%. Heterogeneity was low to moderate (I²=6.2-23.1%). Conclusions: Transcranial ultrasonogra-phy demonstrates good diagnostic accuracy for stroke detection. While operator-dependent and limited by acoustic window availability, these techniques, after thorough training and validation, may be prioritized in low-resource setting were computed tomography is not readily available through the means of formal capacitation and feasibility studies. The moderate risk of bias in half of included studies suggests a lack of higher-quality research.