Abstract A223: Widefield and high-resolution fluorescence imaging of multiple targeted contrast agents to detect dysplasia and cancer in oral tissue

Abstract

Abstract Background: Early detection of dysplasia and cancer in the oral cavity dramatically improves patient survival, however, early stage disease can be difficult to identify with a conventional oral examination. Non-invasive, molecular-specific optical imaging has the potential to rapidly and simultaneously assess expression of multiple biomarkers that indicate neoplastic changes. Imaging biomarker expression may aid not only in early detection, but also in determination of an appropriate surgical margin, ensuring that all neoplastic tissue is removed without removing excess normal tissue. Methods: Three contrast agents were topically applied to freshly resected oral lesions (n=20) for 20 minutes. One contrast agent targets overexpression of epidermal growth factor receptor and is composed of epidermal growth factor peptide labeled with a fluorescent dye. Another is a fluorescently labeled glucose that targets increased metabolism due to increased expression and activity of GLUT transporters. The third intercalates with nucleic acids to allow quantification of metrics such as nuclear size and inter-nuclear distance. After incubation, both widefield and high-resolution fluorescence images were collected for analysis. Widefield imaging increases sensitivity of detection by surveying a large region, while high-resolution imaging increases specificity by interrogating a smaller field of view in more detail. Using an auto-threshold function in Matlab software, the widefield fluorescence images were converted to binary. This binary image was smoothed to identify a contiguous region, which was outlined as a potential “margin.” This potential margin was overlaid onto a pathology map created through intensive review of H&E slides with an expert head and neck pathologist to assess agreement of the predicted margin to the gold standard. Quantitative features from high-resolution images were analyzed in Matlab and used to create an objective classification algorithm to classify images as either normal or neoplastic. Immunohistochemistry was used to confirm biomarker expression. Results: The predicted margin from both contrast agents matches the pathology map well. Not only can invasive cancer be detected, but difficult to identify regions of mild dysplasia are also recognized as abnormal. Varying the threshold adjusts the sensitivity and specificity of this margin, which can be calculated based on the area of the region correctly diagnosed or not. The objective classification algorithm created from high-resolution fluorescence images of the nucleic acid intercalator performed with 85% sensitivity and 78% specificity. IHC confirms expression level of epidermal growth factor receptors and GLUT transporters. Conclusions: We have demonstrated non-invasive topical delivery of contrast agents that may help identify oral neoplasia. Contrast agents can be multiplexed to label many biomarkers associated with oncogenesis. Detection using both widefield and high-resolution imaging permits rapid, objective assessment of biomarker expression at the point of care. This may assist a clinician in choosing a region to biopsy for initial diagnosis or in determining an appropriate surgical margin during treatment. The technique can also be applied to other epithelial cancers, thus impacting a large range of clinical cases. Citation Information: Mol Cancer Ther 2009;8(12 Suppl):A223.