An Optical Imaging Threshold to Detect Head and Neck Cancer during Fluorescence-Guided Surgery
Lindsay Moore, University of Alabama at Birmingham
During fluorescence-guided surgery, a cancer-specific optical probe is injected and then visualized using a compatible device intraoperatively to provide visual contrast between disease and normal tissues. However, the current classifications used to define the extent of fluorescence within tissues are often purely qualitative, undefined, and lacking distinct criteria for fluorescence comparison (Nat Med. 2011;17:1315-9). A quantitative reporting criterion using standardized methods is necessary for widespread approval and advancement of this technique. Here, we introduce a ratiometric value as a novel method for assessing tissue fluorescence in real-time to objectively distinguish between normal and cancerous tissue during fluorescence-guided surgery. Imaging was performed on punch biopsy tissues from resected primary tumors of head and neck cancer during a phase 1 dose-escalation trial evaluating the safety and tumor-specificity of cetuximab-IRDye800. In this study, 97 punch biopsies of tumor (n=50) and normal (n=47) tissue were collected from areas of high and low fluorescence intensity in 12 resected primary tumors from patients who received the study probe. Two fluorescent imaging devices, an open-field (LUNA) and a closed-field (Pearl) device, were used to evaluate the approach and assess the variability of a ratiometric threshold between different imaging devices. Punch biopsy tissues were imaged using both imaging devices and mean fluorescence intensity (MFI) was calculated. Additionally, skin and muscle samples were collected and imaged to serve as internal anatomic controls for each patient and to establish a patient-matched “background” fluorescence to ensure the ratiometric value would account for inherent variability between patient tissues. For each specimen, a fluorescence ratio was calculated by dividing the specimen’s MFI by the MFI of patient-matched background tissues (muscle or skin). Ratios were correlated to pathological assessment in order to determine a ratiometric threshold to predict the presence of malignant tissue with fluorescence imaging. For the open-field intraoperative device, when the muscle-normalized ratiometric value was less than 2.7, the negative predictive value (NPV) was 89.2%. When skin-normalized values were less than 1.1, the NPV was 92.0%. With the closed-field device, the muscle-normalized NPV was 89.8% using a threshold of 3.2 and skin-normalized NPV was 90.7% using a threshold of 1.5. Table 1 shows the sensitivity and specificity of each imaging device to accurately detect disease using a corresponding threshold. Receiver operator characteristic (ROC) analysis was performed to assess the accuracy of the diagnostic test, which demonstrated a significantly (p<0.01) greater AUC than the chance diagonal with a 95% confidence interval for both skin and muscle normalized ratios using both open- and closed-field imaging devices. This proof of concept study yielded a highly sensitive, semi-quantitative threshold for objective determination of the presence of cancer using fluorescence imaging. Furthermore, statistical assessments of the ratiometric threshold confirmed the accuracy of the diagnostic tool.