191 Real Time Assessment of Lymph Nodes and Resection Margins Using Diffuse Reflectance Spectroscopy in Oesophagogastric Cancer

Abstract Background Resection margins and lymph node status in upper gastrointestinal cancer surgery are crucial to postoperative outcomes, but their intraoperative assessment remains challenging. The current gold-standard, frozen section, is time-consuming, operator-dependent and limited to small sampling areas. Diffuse reflectance spectroscopy (DRS) is an optical technique that differentiates tissue by analysing the absorption and reflection of light within tissue. We investigated the performance of DRS in differentiating normal and malignant oesophageal, gastric and lymph node tissue ex- and in-vivo. Method Patients undergoing oesophagectomy or gastrectomy between May 2022 and November 2024 were recruited at Imperial College Healthcare NHS Trust, London. Optical spectra were collected ex-vivo from lymph nodes and in-vivo from normal and suspected tumour tissue in the stomach and oesophagus using a custom-built handheld sterilisable DRS probe. Spectra were correlated with histopathology for ground-truths labelling and used to train an XGB machine learning classifier for binary classification. Performance was evaluated for sensitivity, specificity and accuracy. Results 6,470 stomach and 3,116 oesophageal spectra were acquired from 32 patients in-vivo, and 12,685 spectra were acquired from 169 lymph nodes from 45 patients ex-vivo. For differentiating normal from cancer tissue, XGB achieved sensitivity, specificity and diagnostic accuracy as follows: stomach—77%, 95% and 90%; oesophagus—85%, 81% and 83%; and lymph nodes—90%, 98% and 96%. Conclusions DRS combined with real-time tracking and machine learning can accurately differentiate normal and malignant tissue in the oesophagus, stomach and lymph nodes. This supports its translational potential for integration in operating theatres to enhance surgical precision and improve cancer outcomes.