Signal-to-Noise Ratio Imaging and Real-Time Sharpening of Tumor Boundaries for Image-Guided Cancer Surgery

Abstract

Fluorescence-guided cancer surgery is of considerable current interest in bioanalytical chemistry, engineering, and medicine, but its clinical utility is still hampered by the diffusive (scattering) nature of human tissues and large variations among different patients. Here, we report a new method based on signal-to-noise (contrast-to-noise) ratio (SNR or CNR) imaging for real-time delineation and sharpening of tumor boundaries during image-guided cancer surgery. In particular, we show that in vivo tumor fluorescence signals (both intensity and standard deviation) are strongly correlated with those of the surrounding tissue of the same tissue type and that this relationship is maintained as a function of time for fluorescent tracers such as indocyanine green. This dynamic relationship permits a precise removal of nonspecific background fluorescence from tumor fluorescence. As a result, single-pixel SNR values have been calculated, mapped, and displayed across a large surgical field at 60 frames per second. Pathological validation studies indicate that these SNR values correspond to statistical confidence levels similar (but not identical) to those of normal distributions. When the tumor fluorescence has an SNR of 3, pathological data show a confidence level of approximately 95% in identifying the true tumor lesions. For clinical relevance, we have also carried out first-in-human clinical studies for both oral and esophageal tumors, achieving tumor margin precisions of 1–2 mm with 87.5% histological accuracy and no false positives.