Artificial intelligence in pituitary surgery: the path to clinical solutions

Abstract

Artificial intelligence (AI) is rapidly moving from conceptual innovation to high-performing algorithms across the pituitary patient pathway, promising benefit to patients, endocrinologists and surgeons alike. Pre-operatively, machine learning applied to facial imaging, and natural language processing of electronic health records, show potential for earlier identification of pituitary adenomas and timelier referral from primary providers to a specialist endocrinologist. Intraoperatively, computer vision systems have already augmented surgical training but have the potential to become integrated real-time decision support systems. Post-operatively, predictive models may help to forecast complications and longer-term remission, endocrinological outcomes or recurrence. Across these domains, the emergence of large, multimodal datasets which integrate clinical text, endocrinological investigations, radiological imaging, and intraoperative video promise further AI improvements, yet the impact of AI on real-world clinical decision making is less clear and depends as much on implementation processes and clinicians themselves. This review provides an overview of AI technologies for pituitary patients and explores some of the challenges translating them to clinical practice.