International Journal of Clinical Case Reports, 2025, Vol.15, No.3, 98-109 http://medscipublisher.com/index.php/ijccr 101 3.3 Datasets and algorithms needed for the development and training of AI models The development of effective AI-assisted diagnostic systems relies heavily on the availability of large, annotated datasets and advanced algorithms. High-quality datasets are essential for training AI models to recognize and interpret medical images accurately. These datasets often include a diverse range of images from different patient populations and imaging modalities, ensuring that the models can generalize well to new data (Hadjiiski et al., 2022). For instance, the Alberta Stroke Program Early CT Score (ASPECTS) dataset is commonly used to train models for stroke detection, utilizing random forest learning (RFL) and CNNs to achieve high sensitivity and specificity (Murray et al., 2019). In addition to datasets, the choice of algorithms plays a crucial role in the performance of AI models. Deep learning algorithms, such as CNNs, have been particularly successful in medical imaging due to their ability to automatically extract relevant features from raw image data (Chan et al., 2015; Murray et al., 2019). Transfer learning, which involves pre-training a model on a large dataset and then fine-tuning it on a smaller, task-specific dataset, is another technique that has proven effective in medical diagnostics (Kermany et al., 2018). This approach allows for the development of robust models even when limited annotated data is available, making it a valuable tool in the field of AI-assisted diagnostics. 4 Application of AI-Assisted Diagnostic Systems in Emergency Care for Cerebrovascular Accidents 4.1 Real-time detection and diagnostic capabilities AI-assisted diagnostic systems have shown significant promise in the real-time detection and diagnosis of cerebrovascular accidents, particularly in the context of acute ischemic strokes. These systems leverage advanced machine learning algorithms, such as convolutional neural networks (CNNs), to rapidly identify large vessel occlusions (LVOs) from imaging data. For instance, the use of CNNs has demonstrated a sensitivity of 85% in detecting LVOs, which is notably higher than the 68% sensitivity achieved by traditional random forest learning methods (Murray et al., 2019). This rapid assessment capability is crucial in emergency settings where timely intervention can significantly impact patient outcomes. Moreover, AI systems like Viz.ai and Brainomix have been integrated into clinical workflows to expedite the detection and treatment of strokes. These platforms automatically analyze imaging data to identify LVOs and activate emergency stroke treatment protocols, thereby reducing the time to treatment and improving the chances of patient recovery (Murray et al., 2019; Gilotra et al., 2023). The ability of these systems to provide real-time, accurate diagnoses helps mitigate the delays and variability associated with human interpretation, ultimately enhancing the efficiency of emergency care for stroke patients. 4.2 Clinical trials and case studies Several clinical trials and case studies have been conducted to evaluate the effectiveness of AI-assisted diagnostic systems in improving stroke care. The GOLDEN BRIDGE II trial, for example, is a multicenter, cluster-randomized study in China that investigates the impact of an AI-based clinical decision support system (AI-CDSS) on stroke outcomes. This trial involves 80 hospitals and aims to assess whether the AI-CDSS can reduce the incidence of new vascular events by providing AI-assisted imaging analysis and guideline-based treatment recommendations (Li et al., 2023). Preliminary results suggest that the AI-CDSS could significantly enhance stroke care quality and patient outcomes. In another study, the real-world performance of the Viz LVO software was evaluated across a multihospital stroke network. The study found that the AI software achieved a sensitivity of 93.8% and a specificity of 91.1% in detecting LVOs, demonstrating its utility as an adjunct tool in stroke diagnostics (Matsoukas et al., 2022). These findings underscore the potential of AI systems to improve diagnostic accuracy and streamline the management of cerebrovascular accidents in emergency settings. 4.3 Diagnostic accuracy of AI systems Comparative studies have highlighted the superior diagnostic accuracy of AI systems over traditional methods in detecting cerebrovascular accidents. For instance, a systematic review of AI applications in stroke diagnosis
RkJQdWJsaXNoZXIy MjQ4ODYzNA==