Abstract
RATIONALE: Lung carcinoma diagnosis on tissue biopsy can be challenging because of insufficient tumor and lack of architectural information. Optical coherence tomography (OCT) is a high-resolution imaging modality that visualizes tissue microarchitecture in volumes orders of magnitude larger than biopsy. It has been proposed that OCT could potentially replace tissue biopsy. OBJECTIVES: We aim to determine whether OCT could replace histology in diagnosing lung carcinomas. We develop and validate OCT interpretation criteria for common primary lung carcinomas: adenocarcinoma, squamous cell carcinoma (SCC), and poorly differentiated carcinoma. METHODS: A total of 82 ex vivo tumor samples were included in a blinded assessment with 3 independent readers. Readers were trained on the OCT criteria, and applied these criteria to diagnose adenocarcinoma, SCC, or poorly differentiated carcinoma in an OCT validation dataset. After a 7-month period, the readers repeated the training and validation dataset interpretation. An independent pathologist reviewed corresponding histology. MEASUREMENTS AND MAIN RESULTS: The average accuracy achieved by the readers was 82.6% (range, 73.7-94.7%). The sensitivity and specificity for adenocarcinoma were 80.3% (65.7-91.4%) and 88.6% (80.5-97.6%), respectively. The sensitivity and specificity for SCC were 83.3% (70.0-100.0%) and 87.0% (75.0-96.5%), respectively. The sensitivity and specificity for poorly differentiated carcinoma were 85.7% (81.0-95.2%) and 97.6% (92.9-100.0%), respectively. CONCLUSIONS: Although these results are encouraging, they indicate that OCT cannot replace histology in the diagnosis of lung carcinomas. However, OCT has potential to aid in diagnosing lung carcinomas as a complement to tissue biopsy, particularly when insufficient tissue is available for pathology assessment.