New optical imaging techniques have been developed in order to improve diagnostics precision. Optical coherence tomography (OCT) is based on low-coherence interferometry to detect the intensity of backscattered infrared light from biological tissues by measuring the optical path length. OCT provides the advantage of real-time, in vivo, low-cost imaging of suspicious lesions without having to proceed directly to a tissue biopsy.
The post-processing techniques can be used for improving the precision of diagnostics and providing solutions to overcome limitations for OCT. Image processing can include noise filtration and evaluation of textural, geometric, morphological, spectral, statistic and other features. The main idea of this investigation is using information received from multiple analyze on 2D- and 3D-OCT images for skin tumors differentiating.
At first, we tested the computer algorithm on OCT data hypercubes and separated B- and C-scans. Combination of 2D and 3D data give us an opportunity to receive common information about tumor (geometric and morphological characteristics) and use more powerful algorithms for features evaluation (fractal and textural) on these separated scans. These groups of features provide closer connection to classical wide-used ABCDE criteria (Asymmetry, Border irregularity, Color, Diameter, Evolution). We used a set of features consisting of fractal dimension, Haralick's, Gabor's, Tamura's, Markov random fields, geometric features and many others.
We could note about good results on the test sets in differentiation between BCC and Nevus, MM and Healthy Skin. We received dividing MM from Healthy Skin with sensitivity more 90% and specificity more 92% (168 B-scans from 8 species) by using three Haralick’s features like Contrast, Correlation and Energy. The results are very promising to be tested for new cases and new bigger sets of OCT images.