Automatic differentiation of nonkeratinized stratified squamous epithelia and columnar epithelia through feature structure extraction using OCT

As a type of precancerous lesion, metaplasia is usually considered to be associated with developing cancer. In clinical practice, surveillance of metaplastic cases usually relies on excisional biopsy followed by histological processing and analysis. As it is an invasive method accompanied by other complications, non-invasive imaging methods such as optical coherence tomography (OCT) can complement the existing method by enabling large area scanning. However, because it takes time to review large amount of data acquired from the whole suspected mucosal areas, an automatic classification method is preferred to alleviate the laboring hours and to avoid ‘sampling errors’ during image analysis. In this study, we report an automatic method to differentiate non-keratinized squamous epithelia and columnar epithelia in OCT images. A high detection accuracy is achieved by using feature structure extraction techniques in intact tissues. Agency for Science, Technology and Research (A*STAR) Ministry of Education (MOE) Nanyang Technological University National Research Foundation (NRF) This research was supported in part by A*STAR Biomedical Research330 Council (H1701a008), National Natural Science Foundation of China (Grant331 No. 61705184), National Research Foundation Singapore (NRF-CRP13-332 2014-05), Ministry of Education Singapore (RG 83/18 (2018-T1-001-144)), and 333 NTUAIT-MUV program in advanced biomedical imaging (NAM/15005).