Smart Surgical Margins: An Artificial Intelligence Powered RealTime Cancer Residual Detection Using Fluorescence Imaging System: A Cross-Sectional Study at a Tertiary Care Hospital in Mandya

This is a prospective observational cross sectional diagnostic accuracy study that evaluates an ex vivo near infrared indocyanine green fluorescence imaging system with computer vision and machine learning for assessment of surgical margins on excised breast and colorectal cancer specimens. Fresh inked specimens are imaged in a light tight enclosure under 785 nm excitation with about 830 nm emission capture. The index test produces margin heatmaps and tumour to background ratio metrics on an NVIDIA Jetson platform. The reference standard is histopathology H and E of targeted and control slices from each margin face. Primary outcomes are sensitivity specificity and area under the ROC curve for face level classification against histopathology. Secondary outcomes include time to result per slice diagnostic metrics inter observer agreement and concordance of AI hotspots with histologic margin distance. Adults of eighteen years and above undergoing curative intent surgery are included. Metastatic disease and specimens unsuitable for imaging are excluded. The goal is to deliver same day triage that guides pathology slicing reduces delays and builds a labelled dataset for future in vivo translation.