Feasibility of diffuse reflectance spectroscopy for tumor border detection in colorectal cancer surgery

Adequate distal resection margins (DRM) in colorectal cancer surgery are critical for oncological safety while preserving sphincter function. Current intra-operative DRM assessment methods are limited or have major drawbacks, leaving surgeons without real-time guidance. Diffuse reflectance spectroscopy (DRS) offers a potential solution using rapid optical feedback on tissue characteristics. In this study, continuous measurements were performed on freshly excised colorectal specimens using a broadband DRS system (400–1600 nm). A binary classification approach was applied to discriminate between non-tumor and tumor locations. Multiple feature extraction strategies were evaluated in combination with different classifiers and hyperparameter tuning was performed. A total of 39,741 spectra (22,451 non-tumor, 17,290 tumor) from 86 patients were included. Using the first derivative of the spectra paired with a radial basis function (RBF) support vector machine achieved good results, reaching a mean Matthews Correlation Coefficient (MCC) of 0.76 ± 0.27 across patients. Sensitivity and specificity were 0.93 ± 0.13 and 0.88 ± 0.17, respectively. This study demonstrates the feasibility of using DRS combined with machine learning to distinguish tumor from non-tumor regions in colorectal cancer surgery. Future work should focus on prospective validation of distal margin assessment and exploring probe designs with deeper penetration to enhance intra-operative applicability.