BACKGROUND AND OBJECTIVES: Middle meningeal artery (MMA) embolization is an emerging treatment option for chronic subdural hematoma. Surgeons must pay close attention to multiple vessels when using liquid embolic agents to avoid complications. Unintended embolization through dangerous anastomotic connections can result in serious complications, such as visual loss or cranial nerve dysfunction. In this study, we report our preliminary experience with real-time artificial intelligence (AI)–assisted MMA embolization and evaluate its performance. METHODS: An AI-based system (iMed Technologies) was used for 19 lesions in 15 patients at 2 institutions, with 4 patients receiving bilateral treatment. The software automatically detects liquid embolic agents in biplane fluoroscopy images in real-time and notifies operators when the agent reaches any of the predefined areas. The safety, efficacy, and accuracy of the notifications were retrospectively evaluated using recorded videos. RESULTS: A total of 36 MMA branches were embolized using n-Butyl-2-cyanoacrylate. The mean true positives, false negatives, and false-positive notifications per vessel embolization were 8.9, 1.6, and 0.9, respectively. The precision and recall of the notifications were 90.9% and 84.9%, respectively. In 40.2% of the true-positive notifications, operators immediately paused agent injection after receiving the notification, demonstrating potential clinical effectiveness of the AI system. No adverse events were reported. CONCLUSION: To our knowledge, this is the first study to examine MMA embolization for chronic subdural hematoma with real-time AI assistance. The system demonstrated high notification accuracy, safety, and potential clinical usefulness for liquid embolization procedures. Large-scale prospective studies are warranted to validate the impact on clinical outcomes.
Kono et al. (2025) studied this question.