High–Speed Ultrahigh–Resolution Optical Coherence Tomography (SOCT) Spectral Retinal Oximetry

Abstract

Purpose: : To determine if the spectral data obtained using SOCT are suitable for measurements of retinal blood oxygen saturation.

Methods: : Three–dimensional disc–centered retinal tissue volumes (501 x 180 x 2048 pixels; 6mm x 6mm x 1.5mm) were assessed in 15 normal healthy subjects by SOCT. Oximetry Measurement: An enface image was created by integrating total reflectance for every A–scan. Artery (AR) and vein (VN) pairs were located and specific A–scans identified for analysis. The criteria for A–scan inclusion were (1) locations at vessel edges, avoiding the specular reflection that occurs near the center of vessels, and (2) avoidance of tissue locations with the appearance of light saturation in the enface image. Ten A–scans each from the edge of an AR, adjacent tissue, and the edge of a VN were sampled. The model for light interaction with blood used previously in photographic retinal oximetry was applied to the current analysis. Using this model, smaller optical density ratios (ODRs) represent higher saturation levels. Arterial and venous ODR were; therefore, calculated as: ODR_AR = ln(Tissue855 /AR 855) / ln (Tissue 805 / AR 805) ODR_VN = ln(Tissue855 /VN 855) / ln (Tissue 805 / VN 805) with Tissue, AR, and VN representing total reflectance within an SOCT A–scan sampled at the 805nm or 855nm centered bandwidth. AR and VN ODRs were compared by Wilcoxon signed rank test, with p < 0.05 considered significant.

Results: : In all subjects, AR ODRs were less than VN ODRs, representing high levels of oxygen saturation in the AR samples. However, there was a large inter–individual variability in AR and VN ODRs. [AR ODRs –2.667 to 1.427, VN ODRs from –1.469 to 10.298] AR ODRs (–0.303 ± 1.017) were significantly less than VN ODR’s (1.56 ± 2.695). (Mean ± SD, p = 0.0007)

Conclusions: : Application of light absorption and scatter models to a spectral analysis of reflectance levels in SOCT images appears to produce predictable and consistent ODRs; specifically, measurements of oxygen saturation higher in ARs than VNs. Retinal blood oxygen saturation measurements may be obtained from the SOCT image dataset; however, much work remains in the calibration and interpretation of SOCT oximetry data.

Keywords: imaging/image analysis: clinical • image processing • metabolism