Repository of DUTH

The role of artificial intelligence in endoscopic skull base surgery (Master thesis)

Γκότσης, Αντώνιος/ Gkotsis, Antonios/ Ρινολογία - ρινοχειρουργική

Background: Artificial intelligence (AI) is emerging as an important technological tool in modern surgery, with applications ranging from medical image analysis and preoperative planning to intraoperative navigation, assessment of surgical performance, and surgical data analysis. Endoscopic skull base surgery is particularly suitable for the implementation of these technologies because of the complex anatomy of the region, its close relationship with critical neurovascular structures, and the continuous generation of imaging and intraoperative data. Aim: The aim of this thesis was to review and critically appraise the available literature on the role of AI in endoscopic skull base surgery, with emphasis on preoperative planning, intraoperative guidance, endoscopic transsphenoidal pituitary surgery, augmented reality, robotic surgery, and surgical data science. Methods: A narrative literature review was conducted, using PubMed as the primary database and Scopus and Google Scholar as supplementary sources. Original studies, reviews, proof-of-concept studies, and technical studies relevant to the topic were included, with the period 2000–2026 serving as the primary timeframe. The selected literature was evaluated and synthesized thematically according to the main applications of AI in skull base surgery. Results: Among the applications examined, medical image analysis, automated segmentation, and recognition of anatomical structures show the greatest degree of technological development. Considerable research interest has also focused on endoscopic video analysis, automated registration of imaging data, and intraoperative support systems. In transsphenoidal pituitary surgery, machine-learning and deeplearning models have been investigated for both preoperative planning and prediction of surgical outcomes. In parallel, digital surgery, augmented reality, surgical data science, and collaborative robotics represent related technological fields that may incorporate AI-based applications. However, widespread clinical integration remains limited by data heterogeneity, small datasets, limited external validation, and issues related to interpretability, human factors, protection of personal data, and medical liability. Conclusions: AI may enhance preoperative planning, intraoperative guidance, and evaluation of the surgical process, acting as a supportive technology rather than replacing the surgeon. Its future clinical value will depend not only on algorithmic performance, but also on safe integration into the surgical workflow, user acceptance, and demonstration of benefit in clinically meaningful patient-relevant outcomes.