Back to 2025 Abstracts

---

IN VIVO REAL-TIME ASSESSMENT OF RESECTION MARGINS USING DIFFUSE REFLECTANCE SPECTROSCOPY IN UPPER GASTROINTESTINAL CANCERS
Maxime Giot^*, Naim Slim, Scarlet Nazarian, Ioannis Gkouzionis, Ria Ranjitkar, Priscilla Anketell, Robert Goldin, Jo Lloyd, Christopher Peters, Daniel Elson
Imperial College London, London, England, United Kingdom

Background
Accurate assessment of resection margins in surgical resection of upper gastrointestinal cancers is crucial to ensure patient survival and preventing recurrence. However, differentiating between normal, fibrotic, and tumor tissue intraoperatively accurately and without lengthening operation time remains a significant challenge due to the lack of a real-time tissue differentiation technique. This challenge is further complicated by factors such as tissue scarring or fibrosis caused by neoadjuvant therapies, which can make tissue differentiation more difficult.
Diffuse reflectance spectroscopy (DRS) is a point-based optical sensing technique which can distinguish tissue types by analyzing the absorbance and diffuse reflection of light in the tissue sampled. The aim of this study was to assess the diagnostic accuracy of a DRS probe integrated with a real-time tracking system to differentiate tissue types in vivo, with the aim of enhancing surgical precision.

Methods
Patients undergoing esophagectomy or gastrectomy for upper gastrointestinal adenocarcinoma were recruited between May 2022 and October 2024. Optical spectra were collected intraoperatively from normal and suspected tumor tissue in the stomach and esophagus using a custom-built, sterilizable DRS probe with a tracking system. The spectra were correlated with histopathology analysis to establish ground truth labels. An extreme gradient boosting (XGB) machine learning classifier was trained and tested to classify spectra as normal or tumor. Classifier performance was evaluated for sensitivity, specificity, accuracy, and area under the receiver operating characteristic (ROC) curve (AUC).

Results
A total of 9,954 spectra were acquired from 34 patients, including 6,838 spectra from stomach tissue and 3,116 from esophageal tissue. For stomach tissue, the XGB classifier achieved a sensitivity of 67%, specificity of 90%, diagnostic accuracy of 84%, and an AUC of 0.91. For esophageal tissue, sensitivity and specificity were 80% and 73%, respectively, with an accuracy of 78% and an AUC of 0.85.

Conclusion
These findings demonstrate that DRS, combined with real-time tracking and machine learning, can accurately differentiate normal and tumor esophageal and gastric tissues in vivo. Unlike the gold standard of frozen sections, which presents several limitations –sensitivity of 67%, taking on average 30 minutes, operator dependency and the inability to assess large tissue areas –, DRS provides real-time, objective tissue differentiation with high accuracy. This study supports the translational potential of integrating DRS into the operating room as a reliable and efficient alternative to frozen sections, with the potential to enhance surgical precision, reduce rates of positive resection margins, and ultimately improve outcomes for cancer patients.


Figure 1 (a) Schematic diagram of the setup of the DRS instrumentation. (b) Photograph of the DRS probe being used in vivo.

Figure 2. Real-time tracking of the DRS probe tip during in vivo scanning of tissue with an overlay image of the sampled areas with green points.

---

Back to 2025 Abstracts