DR P. TARUN VARMA,DR NANDITHA GUDI,DR RAJESH S.
DOI: https://doi.org/Background:Recent advancements in artificial intelligence (AI) are revolutionizing the healthcare landscape, particularly within the realm of radiology. For patients involved in road traffic accidents (RTAs), the capability to rapidly and accurately diagnose intracranial hemorrhage (ICH) is of utmost importance, as even a slight delay in diagnosis can have devastating consequences on morbidity and mortality rates. This study delves into the potential of AI, specifically through the application of Convolutional Neural Networks (CNNs), to swiftly and effectively identify ICH in CT scans. Furthermore, we aim to establish how the diagnostic accuracy of AI — utilizing the powerful ResNet-50 model — compares to that of seasoned radiologists.
Methods: A cross-sectional study was conducted at Saveetha Medical College and Hospital. CT brain images of RTA patients were collected. The open-source CNN model ResNet-50 was applied to classify the presence or absence of ICH. Performance was evaluated using sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).
Results: ResNet-50 achieved a sensitivity of 90.4% and a specificity of 92.8%. The PPV was 96.63%, and the NPV was 90.63%.
Conclusion: CNNs like ResNet-50 show promise in detecting ICH and could complement radiologists in clinical practice, enhancing efficiency and reducing diagnostic delays. Further validation with larger datasets and real-time clinical integration is necessary before widespread adoption.
