Oxygen saturation in optic nerve head structures by hyperspectral image analysis

A method is presented for the calculation and visualization of percent blood oxygen saturation from specific tissue structures in hyperspectral images of the optic nerve head (ONH).

Trans-pupillary images of the primate optic nerve head and overlying retinal blood vessels were obtained with a hyperspectral imaging (HSI) system attached to a fundus camera. Images were recorded during normal blood flow and after partially interrupting flow to the ONH and retinal circulation by elevation of the intraocular pressure (IOP) from 10 mmHg to 55 mmHg in steps. Percent oxygen saturation was calculated from groups of pixels associated with separate tissue structures, using a linear least-squares curve fit of the recorded hemoglobin spectrum to reference spectra obtained from fully oxygenated and deoxygenated red cell suspensions. Color maps of saturation were obtained from a new algorithm that enables comparison of oxygen saturation from large vessels and tissue areas in hyperspectral images.

Percent saturation in retinal vessels and from the average over ONH structures (IOP = 10 mmHg) was (mean +/- SE): artery 81.8 +/- 0.4%, vein 42.6 +/- 0.9%, average ONH 68.3 +/- 0.4%. Raising IOP from 10 mmHg to 55 mmHg for 5 min caused blood oxygen saturation to decrease (mean +/- SE): artery 46.1 +/- 6.2%, vein 36.1 +/- 1.6%, average ONH 41.9 +/- 1.6%. The temporal cup showed the highest saturation at low and high IOP (77.3 +/- 1.0% and 60.1 +/- 4.0%) and the least reduction in saturation at high IOP (22.3%) compared with that of the average ONH (38.6%). A linear relationship was found between saturation indices obtained from the algorithm and percent saturation values obtained by spectral curve fits to calibrated red cell samples.

Percent oxygen saturation was determined from hyperspectral images of the ONH tissue and retinal vessels overlying the ONH at normal and elevated IOP. Pressure elevation was shown to reduce blood oxygen saturation in vessels and ONH structures, with the smallest reduction in the ONH observed in the temporal cup. IOP-induced saturation changes were visualized in color maps using an algorithm that follows saturation-dependent changes in the blood spectrum and blood volume differences across tissue. Reduced arterial saturation at high IOP may have resulted from a flow-dependent mechanism.