Multi-Spectral Clinical Prototype for Fluorophore Detection

Abstract

Identification of tumour margins during resection of the brain is critical for improving the post-operative outcomes. Current methods of tumour identification use 5-ALA, an exogenous precursor, metabolized to fluorescent PpIX in tumour tissue. Although visible under fluorescent microscope, PpIX is easily photo-bleached and tumour tagging is subjective, resulting in tumour under-resection and accelerated recurrence. To address this issue, photo-bleaching resistant and quantitative method is required. This study describes the characterization of a pulsed, multi-wavelengths system designed to measure diffuse reflectance and auto-fluorescence under strong ambient illumination conditions. The performance was tested on n = 400 liquid tissue phantoms containing a wide concentration range of absorber, scatterer and two fluorophores as well as on ex-vivo samples of gray and white matter. The background subtraction technique was shown to be efficient for a range of ambient illumination intensities. A linear relationship was observed between system response and predicted fluorophore concentrations as well as 97.8% accuracy of tissue classification by 5-fold cross-correlation, linear SVM.