Ex Vivo Detection of Pre-Cancerous Colorectal Polyps via Optical Coherence Tomography and Deep Learning Methods

Colonoscopy has been very successful in reducing the morbidity and mortality of colorectal cancer. However, it incurs a significant resource and financial burden on the health care system. To address this issue, a leave-in-situ approach for diminutive polyps (≤ 5mm) is under consideration. This study utilized Optical Coherence Tomography (OCT) to distinguish colorectal polyps as either benign (normal and hyperplastic) or malignant potential (adenomatous and sessile serrated adenomatous), achieving diagnostic accuracy appropriate for clinical screening. Three-dimensional OCT image volumes were acquired from polyps from 225 patients and en face images were created. The en face images were annotated by an expert histopathologist. Additional datasets were created by extracting various features from the images and OCT spectral content. The resulting datasets were used to train individual U-Net models to identify malignant potential. A two-stage fusion process was applied, aggregating the classification scores from multiple processed datasets, using a range of traditional machine learning classifiers, to achieve the final decision. The proposed method yielded an accuracy of 90.3%, with sensitivity and specificity values of 96.9% and 81.5%, respectively. The corresponding area under the curve (AUC) was 0.90. These findings demonstrate that OCT imaging of colon polyps holds promise as a complementary tool to colonoscopy, potentially supporting leave-in-situ management approaches.