O-OGC03 Real-time tracking and classification of tumour and non-tumour tissue in upper gastrointestinal cancer specimens using diffuse reflectance spectroscopy
Abstract Background Diffuse reflectance spectroscopy (DRS) is a technique that allows discrimination of normal and abnormal tissue based on spectral data. It is a promising technique for cancer margin assessment. However, application in a clinical setting is limited by the inability of DRS to mark the tissue that has been scanned and its lack of continuous real-time spectral measurements. This aim of this study was to develop a real-time tracking system to enable localisation of the tip of a handheld DRS probe to aid classification of tumour and non-tumour tissue. Methods A coloured marker was attached to the DRS fibre probe and was detected using colour segmentation. A Kalman filter was used to estimate the probe’s tip position during scanning of the tissue specimen. In this way, the system was robust to partial occlusion allowing real-time detection and tracking. Supervised classification algorithms were used for the discrimination between tumour and non-tumour tissue, and evaluated in terms of overall accuracy, sensitivity, specificity, and the area under the curve (AUC). A live augmented view with all the tracked and classified optical biopsy sites were presented, providing visual feedback to the surgeons. Results A green coloured marker was successfully used to track the DRS probe. The measured root mean square error of probe tip tracking was 1.18±0.58mm and 1.05±0.28mm for the X and Y directions, respectively, whilst the maximum measured error was 1.76mm. Overall, 47 distinct sets of tumour and non-tumour tissue data were recorded through real-time tracking of ex vivo oesophageal and gastric tissue. The overall diagnostic accuracy of the system to classify tumour and non-tumour tissue in real-time was 94% for stomach and 96% for the oesophagus. Conclusions We have been able to successfully develop a real-time tracking system for a DRS probe when used on stomach and oesophageal tissue for tumour detection, and the accuracy derived demonstrates the strength and clinical value of the technique. The method allows real-time tracking and classification with short data acquisition time to aid margin assessment in cancer resection surgery.
