.png)
SDSC’s Data Science Roundtable with Dr. Danyal Khan – neurosurgical resident at the National Hospital for Neurology and Neurosurgery in London, UK and PhD student at University College
London (UCL) – explored how computer vision and artificial intelligence (AI) can be meaningfully translated into neurosurgical practice. Dr. Khan reflected on nearly a decade of work applying machine learning to surgical video, and his progression from early step-recognition models to real-time anatomical identification tools in the operating room.
He outlined the clinical challenges of pituitary surgery, including anatomical complexity, high-risk neighbouring structures, and the wide variation in patient outcomes across centers in the UK. His work started by defining the consensus standards for surgical steps, anatomy, and outcomes, so that operative videos could be annotated appropriately and at scale. Initial AI models were then used to generate performance metrics that helped cultivate a culture of open reflection and peer-coaching, which resulted in measurable improvements to surgical performance and patient outcomes.
More recent efforts have shifted towards real-time intraoperative support, in particular, real-time anatomical segmentation and navigation. And a key lesson was the importance of surgeon-centric design. Through iterative feedback, mock operating environments, and human factors research, his team learned that surgeons prefer minimal, controllable interfaces with clear indications of model confidence.
The discussion also highlighted lessons in team building, the importance of curated clinical datasets, close clinician-engineer collaboration, and continuous qualitative feedback, all of which are crucial for successful clinical translation.
