Purchase this article with an account.
A. Miyazawa, M. Yamanari, S. Makita, M. Miura, K. Kawana, T. Oshika, Y. Yasuno; Discrimination of Conjunctiva and Sclera by Analysis of Local Statistics of Polarization Sensitive Optical Coherence Tomography Images. Invest. Ophthalmol. Vis. Sci. 2009;50(13):5671.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Conjunctiva and sclera are not always discriminated in anterior eye optical coherence tomography (OCT), although they have distinctive tissue properties. In contrast, characteristic pattern can be observed at the sclera in a phase retardation OCT measured by polarization sensitive OCT (PS-OCT). This is because sclera consists of collagen and it has birefringence. We demonstrate a new algorithm which discriminates conjunctiva and sclera based on the local statistics of intensity and phase retardation OCT images.
4 eyes of 4 normal subjects were examined by PS-OCT. Local birefringence was calculated with the kernel thickness of 109 um based on measured Jones matrices. Local mean of intensity (I) and birefringence (BR) within a moving window (39 um x 38 um) were calculated for each pixel. The I and BR (referred to as features) were used to discriminate tissue types. The sets of features, pairs of I and BR, which are referred to as feature vectors, were obtained in 3 reference regions for target tissues (conjunctiva, sclera and uvea). Then the similarities to the target tissues were evaluated for each pixel by using the distances from the pixel to the target tissues in the feature vector space. The similarities to conjunctiva, sclera and uvea were then assigned to red, blue, and green color channels of the final image, respectively.
In 4 of 4 cases, sclera and conjunctiva were clearly discriminated as shown in the figure. Although the specificity is low, the part of uvea was discriminated from the sclera and conjunctiva. It was found that the blue region had higher BR than red region. This corresponds to the existence of collagen fibers in sclera.
Conjunctiva and sclera were discriminated by analysis of local statistics of PS-OCT images. This new algorithm is useful to discriminate tissues of PS-OCT images not by anatomical tissue structures but by direct tissue properties.
This PDF is available to Subscribers Only