Optical Coherence Tomography

Publication date: Available online 23 November 2017
Source:Academic Radiology
Author(s): Richard Ha, Lauren C. Friedlander, Hanina Hibshoosh, Christine Hendon, Sheldon Feldman, Soojin Ahn, Hank Schmidt, Margaret K. Akens, MaryAnn Fitzmaurice, Brian C. Wilson, Victoria L. Mango
Rationale and ObjectivesThis study aimed to assess whether different breast cancer subspecialty physicians can be trained to distinguish non-suspicious from suspicious areas of post-lumpectomy specimen margin in patients with breast cancer using optical coherence tomography (OCT) images (a near-infrared based imaging technique) with final histology as the reference standard.Materials and MethodsThis institutional review board-exempt, Health Insurance Portability and Accountability Act-compliant study was performed on 63 surgically excised breast specimens from 35 female patients, creating a 90-case atlas containing both non-suspicious and suspicious areas for cancer. OCT images of the specimens were performed, providing 6.5–15 µm resolution with tissue visualization 1–2 mm subsurface. From the 90-case atlas, 40 cases were chosen for training and 40 were randomly selected for reader assessment. Three breast imaging radiologists, two pathologists, two breast surgeons, and one non-clinical reader were trained and assessed for ability to distinguish non-suspicious from suspicious findings blinded to clinical data and corresponding histology slides. Duration of training and assessment, sensitivity, specificity, positive predictive value, negative predictive value, and the area under the curve for each reader were calculated as well as averages by subspecialty.ResultsThe average training time was 3.4 hours (standard deviation, 1.2). The average assessment time was 1.9 hours (standard deviation, 0.7). The overall average reader sensitivity, specificity, and accuracy for detecting suspicious findings with histologic confirmation of cancer at the surgical margin for all eight readers were 80%, 87%, and 87%, respectively. Radiologists demonstrated the highest average among the disciplines, 85%, 93%, and 94%, followed by pathologists, 79%, 90%, and 84%, and surgeons, 76%, 84%, and 82% respectively.ConclusionsWith relatively short training (3.4 hours), readers from different medical specialties were able to distinguish suspicious from non-suspicious OCT imaging findings in ex vivo breast tissue as confirmed by histology. These results support the potential of OCT as a real-time intraoperative tool for post-lumpectomy specimen margin assessment.



http://ift.tt/2iLpr8D