Bahram Khoobehi and James M. Beach

4.3. Experiment Two

4.3.1. Methods and Materials

4.3.1.1.Animals, anesthesia, blood pressure, and IOP perturbation

Five normal cynomolgus monkeys, 4–4.5 years of age and 2.5–3 kg body weight were used. The animals were housed in an air-conditioned room (22 ± 1◦C and 66 ± 3% humidity) with a 12-hr light–dark diurnal cycle and access to food and water ad libitum. Monkeys were anesthetized with intramuscular ketamine (7–10 mg/kg) with xylazine (0.6–1 mg/kg) and intravenous pentobarbital (25–30 mg/kg). Administration of the anesthetics was repeated alternately every 30 min as required to maintain the animal in deep, stage IV anesthesia, as monitored by blood pressure and heart rate. Prior to IOP elevation, a topical anesthetic was given (proparacaine hydrochloride ophthalmic solution, 0.5%; Alcon, Fort Worth, TX, USA). A veterinary blood pressure monitor with a 5-cm pediatric cuff (model 9301Vl; CAS Medical Systems, Branford, CT, USA) was used to record blood pressure every five minutes throughout an imaging session. One eye was dilated and a 27-gauge needle connected to a saline manometer was inserted into the anterior chamber under slit-lamp examination. IOP was controlled by altering the height of the reservoir and measured by means of a tonometer (Tonopen XL; Medtronic, Jacksonville, FL, USA).

4.3.1.2. Hyperspectral recordings

HSI was done as previously described.

4.3.1.3. Spectral determinant of percentage oxygen saturation

Begin paragraph with sentence: Percent oxygen saturation was found from HSI images using curve fitting methods, as previously described.73 Before performing the curve fit, the recorded spectrum was transformed by the method of Hammer et al.66 to remove influences of nonhemoglobin light absorption and light scattering. This transformation corrects the recorded curves at three isosbestic wavelengths, 522 nm, 569 nm, and 586 nm, in order to match reference curves of oxygenated and deoxygenated blood.

148

Hyperspectral Image Analysis for Oxygen Saturation Automated Localization of the Eye

We modified the transformation to use the three more closely spaced wavelengths at 522 nm, 548 nm, and 569 nm. Saturation was measured at 561 nm, which is a maximum in the difference spectrum. We used this procedure to test the calibration of red cell suspensions. We used the same procedure for in vivo spectra at different IOPs, except, here, we determined saturation by least-squares curve fits to oxygenated and deoxygenated reference curves from red-cell suspensions, containing 25 equispaced wavelengths between 522 nm and 569 nm. Curve fits were performed with a Windows software package (MathGrapher 2.0; Springfield Holding b.v., Noordwijk, the Netherlands). Reference spectra of saturated (Ssat) and desaturated (Sdesat) red-cell suspensions were fit to transformed retinal blood spectra (S) using fitting parameters A and B with an additive term (C), as in Eq. (4.9):

Percentage oxygen saturation was determined by expressing fitting parameters as in Eq. (4.10):

where A and B correspond with best-fit coefficients for oxyhemoglobin and deoxyhemoglobin contributions as defined by Eq. (4.9).

4.3.1.4.Spatial mapping of oxygen saturation: a modification of the previous mapping algorithm incorporating a correction for blood volume

By following the sign and magnitude of the area constructed between the spectral curve and the three-line segments connecting oxygen-insensitive wavelengths (isosbestic points), an oxygen-sensitive index was obtained. Individual oxygen-sensitive areas were normalized for a total reflected intensity by division by the polygonal areas A1, A2, and A3 under each line segment. The OSC (shown in Fig. 4.2) of the algorithm is given by Eq. (4.11):

the value of which increases with saturation. OSC, as defined above, depends on the volume of blood in the recording. Because significantly different blood volume densities exist in vessels and tissue, this difference must be

149

Bahram Khoobehi and James M. Beach

accounted for in order for comparisons from the different structures to be valid. Optic disk blood volumes have previously been found by reflectometry at three wavelengths.69 We used here the area between the hemoglobin absorption band and a line segment connecting the first and last isosbestic points (see Fig. 4.8) to estimate blood volume. Although this area slightly underestimates the true light absorption from hemoglobin over these wavelengths, it varies directly with the change in blood volume. This quantity is normalized by total light intensity from the area under the line to give the BVC in Eq. (4.12),

where b is the area between the spectral curve and the line segment under the curve and B is the area under the line segment. The volume-corrected, RSIs independent of hemoglobin concentration is given in Eq. (4.13):

Saturation maps were constructed by applying the algorithm at each image pixel using a MATLAB script. Numerical values of RSIv, representing the relative saturations of separate structures, were determined from averaged pixels (n > 1000) inside the borders of vessels and distinct areas of the ONH. Individual pixel values were assigned to color codes, in the order of blue, cyan, green, yellow, and red, to represent progressively higher saturations. The relationship between RSIv and percentage saturation was found in order to calibrate the saturation map.

4.3.1.5. Preparation and calibration of red blood cell suspensions

Red-cell suspensions were prepared as follows: 50 ml of blood was drawn from the femoral vein of the monkey and separated into equal volumes. The samples were centrifuged at 3000 RPM (13◦C) for 20 min. The fluid was carefully aspirated leaving packed red cells, to which was added an equal volume of isotonic saline. The red cells were then resuspended, and this procedure was repeated three times. After rinsing, one sample was exposed to air (oxygenated sample) while to the second sample was added Na-dithionate, until the suspension turned blue. Percentage

150