Chapter 4
Hyperspectral Image Analysis
for Oxygen Saturation Automated Localization of the Eye
Bahram Khoobehi and James M. Beach
4.1. Introduction to Oxygen in the Retina
Both the retina and optic nerve head (ONH) have a high demand for oxygen, and changes in the supply of oxygen resulting from vascular disease play an important role in retinal and ONH pathology. The development of a non-invasive means of measuring oxygen saturation in the fundus of the human eye would be useful in the diagnosis and monitoring of numerous disorders. For example, the measurement of retinal and ONH oxygen saturation is essential for a better understanding of the relationship among oxygen consumption, activity, and metabolism, information that is vital to our understanding of diabetic retinopathy. In addition, accurate measurement of vascular efficiency in terms of oxygen saturation could be used to detect the very early onset of glaucoma, a disease in which early detection is crucial for effective treatment. Evidence that vascular inefficiency plays an important role in the pathogenesis of glaucomatous optic neuropathy has been accumulated. Because vascular inefficiency in the ONH resulting from various systemic disorders is directly related to blood flow in the ONH, which, in turn, is related to ONH tissue oxygenation, we propose to develop a practical system to evaluate oxygen saturation in the clinical setting using a recent innovation, hyperspectral imaging (HSI). The hyperspectral technique measures spectral changes within the visible and infrared
LSU Eye Center, Louisiana State University Health Sciences Center, New Orleans, LA, USA.
123
Bahram Khoobehi and James M. Beach
(IR) spectra and provides information on the molecular state of hemoglobin. We propose to adapt an existing prototype hyperspectral camera to perform clinically useful measurements of the oxygenation status of the retina and ONH to quantitate the role of hypoxia in ocular vascular disease. More specifically, our proposed experimental instrumentation will accommodate parallel studies of the oxygen saturation response to increased intraocular pressure (IOP) in normal monkeys and in monkeys with progressive stages of induced early phase glaucoma. The stimulus–response relationship is a stress–strain response to perturbations from the normal IOP (approximately 15 mmHg). In the absence of glaucomatous disease, autoregulation is expected to maintain normal volumetric blood flow and, hence, normal oxygenation in the ONH tissue up to a homeostatic threshold. It is presently believed that autoregulation is impaired in glaucomatous disease,1−5 possibly because of anatomical vascular impairment of the retina and the ONH. It would be of considerable interest to determine if the threshold for autoregulation impairment is affected during the pre-onset stages of early phase glaucoma. The ONH has three distinct microcirculations: (1) the surface nerve fiber layer (NFL), which interfaces with the juxtaposed NFL and is predominantly nourished by the retinal arterioles; (2) the prelaminar region of the ONH, which is sandwiched between the surface NFL and the lamina cribrosa; and (3) the lamina cribrosa region, which is generally nourished by centripetal branches arising directly from the short posterior ciliary arteries.4,6−8 With this new approach, we expect to determine how acute changes in IOP alone or in combination with chronic IOP elevation (glaucoma) affect these circulations independently and/or collectively. The proposed studies are motivated by the potential for the clinical application of this innovative technology in the early diagnosis and monitoring of therapy for ocular vascular diseases in which the associated hypoxia may eventually lead to loss of vision.
Oxygen delivery to tissue can be assessed in part by measuring the oxygen saturation of blood. Although a more complete assessment of tissue oxygen delivery requires blood flow measurements, specific ophthalmic disorders have been linked with abnormal blood oxygen saturation.9−13 Several methods for evaluating the retinal oxygen saturation in retinal vessels have been reported based on spectrographic techniques in combination with photography, electro-optic detection, and digital imaging.14−23
124
Hyperspectral Image Analysis for Oxygen Saturation Automated Localization of the Eye
In addition, phosphorescence,24−27 oxygen microelectrodes,28−30 and magnetic resonance imaging31,32 methods have been developed to monitor oxygen tension (PO2) in structures of the retina. More recently, Khoobehi et al. describe a method for mapping relative oxygen saturation in the ONH, based on digital imagery obtained over a range of optical wavelengths by HSI methods.33
4.1.1. Microelectrode Methods
The most effective treatment for diabetic retinopathy is panretinal photocoagulation (PRP), which completely or partially stops neovascularization.34,35 This treatment applies widespread grids of laser photocoagulation to destroy photoreceptors in the peripheral retina to prevent further damage of the central retina.
One hypothesis36−38 states that the excess oxygen from the choroid no longer being consumed by the peripheral photoreceptors39−42 will diffuse into the inner half of the retina. Without this oxygen diffusion, the possible hypoxia in the inner retina may cause capillary loss43 and regression of neovascularization. Most studies for this hypothesis have only measured preretinal oxygen tension, or preretinal PO2, after PRP. According to studies in the cat retina, PO2 levels only increased after 100% O2 breathing, not during air breathing.36,44,45 In rabbits, there was an increase in preretinal PO2 after PRP during air breathing, but the rabbit retina has limited circulation and is metabolically different from the human retina.46,47 Increased preretinal PO2 was also seen in miniature pigs after photocoagulation with48 and without49 venous occlusion.
A second hypothesis is that photocoagulation balances the levels of growth factors promoting or inhibiting angiogenesis.43 This hypothesis is supported by microarray studies of changes in gene expression in the retina. However, the levels of growth factors may be a direct result of changing PO2 levels.9
The long-term effects of PRP on intraretinal PO2 are clinically relevant, yet previous experiments have only made measurements the same day when lesions were made or only measured preretinal PO2, which has more direct access to the oxygen supply. Linsenmeier measured intraretinal PO2 after PRP under normoxic conditions in vascularized retinas using O2-sensitive
125
Bahram Khoobehi and James M. Beach
microelectrodes inserted into the eye. The understanding of the mechanism of PRP may lead to more clinically appropriate treatments50 and his measurement will help to understand PRP. However, it is too invasive for clinical measurement of O2.
4.1.2. Phosphorescence Dye Method
R.D. Shonat and C.E. Riva have continued research with phosphorescence. C.E. Riva uses a less invasive phosphorescence method to measure PO2 in the retinal and choroidal vessels, as well as the microvasculature of the ONH rim. Phosphorescence is the emission of light during a transition from a long-lived, spin-forbidden, and excited triplet state to a ground state.51−55 When a molecule capable of phosphorescence is excited into the triplet state, it may emit phosphorescence or transfer its energy to another molecule without light emission (quenching). In the blood, oxygen is the only significant quenching agent56 with the degree of quenching dependent on the concentration of oxygen near the phosphorescent molecule. Therefore, by measuring the quenching effect, the intravascular PO2 can be determined.
A Pd complex of meso-tetra (4-carboxyphenyl) porphine (Porphyrin Products, Logan, UT) was used as the triplet-state oxygen probe. A Wild Macrozoom microscope with an epifluorescence attachment (Wild-Leitz USA, Malvern, PA) was used for excitation of the probe and collection of the phosphorescence. The phosphorescence resulting from a 45-W xenon flash lamp mounted on a Leitz lamp housing was observed with an intensified CCD camera (Xybion Electronic Systems, San Diego, CA) placed in the image plane of the microscope. The filter block inside the epifluorescence attachment permitted control of the excitation and collection wavelengths. Excitation was through one of two filter combinations, with the bandwidth covering one of the two excitation peaks of the probe. A long-pass cutoff filter with 50% transmission at 630 nm was used for the observation of the phosphorescence.
The phosphorescence lifetime can be calculated from images collected at different delay times after a flash. The microcomputer was programmed to trigger a flash, wait the prescribed delay time, turn on the CCD intensifier for 2.5 ms, digitize the image, and store it. Eight images were averaged for each delay time.
126