Introduction: Resection of head and neck cancer with positive margins is a crucial prognostic factor, yet no reliable intraoperative diagnostic adjuncts exist. Rapid Evaporative Ionisation Mass Spectrometry (REIMS/iKnife) can analyse smoke generated by dissection with electrocautery to provide near instant tissue classification. However, REIMS has never been investigated in head and neck cancer in vivo. We designed an IDEAL stage 2a developmental study, aiming to evaluate the validity of our diagnostic model, and develop a feasible protocol for a subsequent clinical trial. Methods: Previously-irradiated patients undergoing transoral robotic surgery at the Royal Marsden for recurrence were prospectively recruited (REC: 14/EE/0024). Intraoperative REIMS data was captured and synchronised with the video feed. The surgeon was blinded to REIMS data throughout the procedure. Univariate and multivariate analyses were conducted using AMX, R and Python. Results: 7 patients have been recruited to date (target 20). All were male. Median age was 63 (range 58-74). 4/7 had HPV-positive disease. REIMS signal was successfully acquired in 100% of cases, but was less reliable during deeper dissection. Tumour signals were successfully classified, as were normal tissue types such as muscle and mucosa, in over 95% of captured signal. Metabolic species potentially contributing to classification were identified, including phospholipids, glycerophospholipids and ceramides previously associated with carcinogenesis. Conclusions: REIMS near-instantaneously distinguishes head and neck cancer from normal tissue intraoperatively. We are iteratively developing engineering solutions to optimise signal collection. Recruitment is ongoing. An optimised protocol will allow design of a clinical trial to evaluate the impact of REIMS on survival outcomes. Introduction: Resection of head and neck cancer with positive margins is a crucial prognostic factor, yet no reliable intraoperative diagnostic adjuncts exist. Rapid Evaporative Ionisation Mass Spectrometry (REIMS/iKnife) can analyse smoke generated by dissection with electrocautery to provide near instant tissue classification. However, REIMS has never been investigated in head and neck cancer in vivo. We designed an IDEAL stage 2a developmental study, aiming to evaluate the validity of our diagnostic model, and develop a feasible protocol for a subsequent clinical trial. Methods: Previously-irradiated patients undergoing transoral robotic surgery at the Royal Marsden for recurrence were prospectively recruited (REC: 14/EE/0024). Intraoperative REIMS data was captured and synchronised with the video feed. The surgeon was blinded to REIMS data throughout the procedure. Univariate and multivariate analyses were conducted using AMX, R and Python. Results: 7 patients have been recruited to date (target 20). All were male. Median age was 63 (range 58-74). 4/7 had HPV-positive disease. REIMS signal was successfully acquired in 100% of cases, but was less reliable during deeper dissection. Tumour signals were successfully classified, as were normal tissue types such as muscle and mucosa, in over 95% of captured signal. Metabolic species potentially contributing to classification were identified, including phospholipids, glycerophospholipids and ceramides previously associated with carcinogenesis. Conclusions: REIMS near-instantaneously distinguishes head and neck cancer from normal tissue intraoperatively. We are iteratively developing engineering solutions to optimise signal collection. Recruitment is ongoing. An optimised protocol will allow design of a clinical trial to evaluate the impact of REIMS on survival outcomes.