Diagnosis of basal cell carcinoma and benign neoplasms based on texture analysis of optical coherence tomography images
Yuriy I. Surkov1,2, Izabella A. Serebryakova1,2, Yana K. Kuzinova3, Olga M. Konopatskova3, Elina A. Genina1,2, V.V. Tuchin1,2,4; 1Saratov State University, Saratov, Russia, 2Tomsk State University, Tomsk, Russia, 3 Saratov State Medical University named after V.I. Razumovsky, Saratov, Russia, 4 Institute of Fine Mechanics and Control, Federal Research Center "Saratov Scientific Center of the Russian Academy of Sciences", Saratov, Russia
Abstract
In this paper, specific markers of basal cell carcinoma and benign neoplasms of human skin on optical coherence tomography (OCT) images are proposed. A statistical analysis of the first and second order OCT images of the neoplasm was performed. To assess the homogeneity of the neoplasm volume image, the asymmetry, kurtosis, mean and standard deviation of the intensity distribution histogram, second-order entropy, and correlation of neighboring image pixels were calculated. The object of the study was a set of OCT images of 62 neoplasms of 40 volunteers diagnosed with basal cell carcinoma (BCC) and/or a benign neoplasm. The proposed texture markers may be useful for developing methods for diagnosing skin neoplasms in dermato-oncological practice. The developed model for classifying neoplasm types and subtypes of BCC correctly differentiated 75.9% of benign neoplasms from malignant ones, while on average only 4.78% of all BCC subtypes were incorrectly assigned to the group of benign neoplasms. On average, for 71.4% of images of neoplasms of various subtypes BCC was assigned to the corresponding group.
Speaker
Yuriy Surkov I.
Saratov State University, Saratov, Russia
Russia
Discussion
Ask question
