International Journal of Clinical Case Reports, 2025, Vol.15, No.3, 98-109 http://medscipublisher.com/index.php/ijccr 102 revealed that AI algorithms, particularly those using CNNs, have higher sensitivity and specificity in identifying acute ischemic strokes and LVOs compared to conventional diagnostic approaches (Murray et al., 2019; Kundeti et al., 2021). This enhanced accuracy is attributed to the AI's ability to process large volumes of imaging data quickly and identify subtle patterns that may be missed by human observers. Furthermore, a critical evaluation of commercially available AI diagnostic tools for stroke imaging found that these systems not only provide rapid and accurate diagnoses but also reduce the variability associated with human interpretation (Wardlaw et al., 2022). The study emphasized the need for standardized performance assessment metrics to ensure the reliability and clinical applicability of AI diagnostic tools. Overall, the integration of AI into stroke diagnostics holds promise for improving diagnostic accuracy and patient outcomes in emergency care settings. 5 Performance and Reliability Analysis of AI Systems 5.1 Sensitivity and specificity The sensitivity and specificity of AI-assisted diagnostic systems are critical metrics for evaluating their performance in emergency care for cerebrovascular accidents. For instance, an AI model integrated into a scanner for detecting intracranial hemorrhages demonstrated a sensitivity of 98.1% and a specificity of 89.7% in a clinical setting, indicating high diagnostic accuracy (Kiefer et al., 2023). Similarly, a systematic review and meta-analysis of AI algorithms for detecting cerebral aneurysms reported pooled sensitivity and specificity values of 91.2% and 83.5%, respectively, underscoring the potential of AI in enhancing diagnostic precision (Din et al., 2022). Moreover, AI systems have shown promising results in the detection of acute ischemic stroke (AIS). A deep learning-based triage application for AIS on brain MRI achieved a sensitivity of 90% and a specificity of 89%, highlighting its efficacy in emergency room settings (Kim et al., 2023). These high sensitivity and specificity values suggest that AI systems can reliably identify cerebrovascular events, thereby facilitating timely and accurate clinical decision-making. 5.2 System stability and consistency The stability and consistency of AI systems are paramount for their reliable application in emergency care. In a study evaluating an AI tool for detecting large vessel occlusion (LVO) stroke, the system demonstrated consistent performance across different hospital settings, with a sensitivity of 93.8% and a specificity of 91.1% for internal carotid artery terminus and middle cerebral artery occlusions (Matsoukas et al., 2022). This consistency across various conditions indicates the robustness of the AI system in real-world clinical environments. Additionally, the H-system, an AI-based tool for intracerebral hemorrhage (ICH) management, showed high reliability in processing medical text data and generating treatment plans. The system achieved an accuracy of 88.55%, with a sensitivity of 91.83% and a specificity of 85.71%, demonstrating its stable performance in handling emergency cases (Deng et al., 2022). These findings underscore the potential of AI systems to maintain high performance levels under diverse clinical conditions, ensuring dependable support for healthcare professionals. 5.3 Clinical validation data Clinical validation of AI systems is essential to establish their efficacy and reliability in real-world settings. For example, an AI model for detecting intracranial hemorrhages was validated using 435 non-contrast head CT scans, achieving a diagnostic accuracy of 99% and a negative predictive value of 99.7% (Kiefer et al., 2023). This high level of accuracy and predictive value provides strong evidence for the clinical utility of the AI system in emergency care. Furthermore, a study assessing the diagnostic performance of AI software for detecting ICH on non-contrast CT images reported an overall diagnostic accuracy of 93.0%, with a sensitivity of 87.2% and a negative predictive value of 97.8% (Figure 2) (Seyam et al., 2022). These metrics, derived from a large dataset of 4 450 CT examinations, reinforce the reliability of AI systems in clinical practice. The consistent performance of these AI
RkJQdWJsaXNoZXIy MjQ4ODYzNA==