This study focuses on bringing artificial intelligence into the operating room to assist with pituitary tumour surgeries performed through the nose. These procedures are technically demanding, and training new surgeons is often inconsistent. To address this, researchers at the National Hospital for Neurology and Neurosurgery are testing AI systems that "watch" surgical videos in real-time to identify anatomy, instruments, and the specific phase of the operation. The core goal of the prospective trial is to improve education and team coordination without interfering with the surgery itself. The AI displays its analysis on tablets positioned for the surgical residents and nurses, rather than the lead surgeon. This setup allows the team to follow the procedure's progress, key anatomy and anticipate next steps without the surgeon needing to stop and explain. Because hospital internet can be unreliable, the study is prioritizing specialized hardware from NVIDIA that processes data locally. This "edge computing" approach ensures the AI is fast and doesn't require a live cloud connection to function. This trial will assess the device feasibility (IDEAL Stage 1 study, \~6 cases), followed by early safety and system technical refinement (IDEAL 2a study, \~20-30 cases).
Study Type
INTERVENTIONAL
Allocation
NA
Purpose
OTHER
Masking
NONE
Enrollment
30
Live intra-op AI analysis of endoscopic video feed, with output displayed on supplementary monitor
National Hospital for Neurology and Neurosurgery
London, United Kingdom
Feasibility of live AI video analysis
The primary objective of this study is to evaluate the feasibility of the TouchSurgery platform or NVIDIA AGx/IGx based platforms for prospective AI-based surgical video analysis (via observation, validated implementation assessment and human factors questionnaires; and semi-structured interviews of surgical team members).
Time frame: Immediately after the intervention/procedure/surgery
Safety
* observation for operating surgeon distraction: recorded as discrete instances of unplanned disruption of primary surgeon workflow per surgery, as observed by observer from research team * wider surgical team workflow disruption : recorded as discrete instances of unplanned disruption of wider surgical team workflow per surgery, as observed by observer from research team * AI output inaccuracy and volatility: measured via sampling of 3-5x clips (30-60sec at 5fps) during which surgical scene is static (i.e. during routine anatomical verification checks), and calculating DICE scores (vs groundtruth segmentations) for accuracy estimation and DICE/sec for volatility estimatipon. * AI output latency: measured as discrete instances of unacceptably elevated latency (\>200ms) of the AI output display vs the primary direct surgical feed, as observed by observer from research team.
Time frame: Perioperatively/periprocedurally (surgeon distraction, team disruption); and immediately after the intervention/procedure/surgery (output accuracy, volatility and latency)
Educational yield
To evaluate the utility of the platform for educational purposes. Via structured educational yield questionnaire of surgeons involved in each case
Time frame: Immediately after the intervention/procedure/surgery
Surgical outcomes
* Surgical performance vs matched cohort: measured via modified OSATS on independent surgical video review * Surgical outcomes vs matched cohort: measured via comparative analysis of standardised outcome set
Time frame: Through study completion, an average of 1 year
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.