Intraoperative optical breast tissue characterization device for tumor margin assessment.

Abstract

Abstract Abstract #801 Background: Optical spectroscopy can quantify the tissue composition of normal and malignant breast tissues. Malignant changes which may be probed with optical spectroscopy include changes in light absorbing and scattering tissue constituents. Our multi-disciplinary group is seeking to utilize optical technology for the intra-operative assessment of tumor margins in breast-conserving surgery (BCS) due to the high re-excision rate in this patient population (20-70%). In this application, an optical spectral imaging device is used to survey the tumor margins and provide immediate feedback to the surgeon about the presence of residual malignancy. Methods: Under an IRB-approved protocol, we are currently testing our device on patients undergoing a partial mastectomy at Duke University Medical Center to determine its ability to provide rapid, non-destructive, accurate assessments of margin status. Optical images of breast tumor margins were recorded. Select areas on the imaged tissues were inked for closer pathologic review; the diagnosis of these areas were co-registered with their corresponding pixels from the margin images, for site-level analysis. The pathologic status of the imaged margins was collected from standard surgical pathology, for margin-level analysis. A feature extraction algorithm was used to convert the optical images to tissue composition images. Wilcoxon tests were used to detect significant differences between optically-measured tissues, separated by both site- and margin-level pathology. Results and Discussion: To confirm that the optical device is sensitive to tissue composition differences between negative and positive margins, optical spectra were acquired from 177 path-confirmed margin sites in 75 patients, of which 17 sites were confirmed as positive. Significant differences (P<0.05) in tissue composition features were observed between negative and positive margin sites, including parameters related to hemoglobin content and saturation. The ability of the imaging device to predict margin status on a margin level was then tested. Full margin images were collected from 34 margins in 28 patients, of which 15 margins were positive or close (<2mm) on pathologic review. Significant differences (P<0.05) were observed in the images between positive and negative margins; the most significant parameters were related to β-carotene content (found in fatty tissues) and the light-scattering coefficient (related to cellular morphology and DNA content). Cross-validation of a classification algorithm trained on the margin-level optical images resulted in a sensitivity and specificity for prediction of margin status of 80% and 62%, respectively. These preliminary results indicate that the optical imaging device is sensitive to differences between negative and positive breast tumor margins, and has the potential to reduce re-excision rates in BCS. Citation Information: Cancer Res 2009;69(2 Suppl):Abstract nr 801.