IMPROVED VISUALIZATION OF CHOROIDAL BLOOD FLOW AND ABERRANT VASCULAR
STRUCTURES IN THE EYE
There is very little information about the blood flow through capillary plexuses which occurs on the time scale of the cardiac cycle. In part this is because direct visualiza¬ tion of such plexuses usually is technologically difficult or impossible, and most blood flow measurement methodologies require that data be obtained over many cardiac cycles. Moreover, when the capillary plexuses have complex vascular geometries and are fed by many arterioles, the additional problem of sorting-out blood flow distributions arises. One example of a capillary plexus is that found in the cerebral cortex. Another example, of great interest to scientists studying the eye, is the choriocapillaris, one of three blood vessel layers of the choroid.
The choroidal circulation of the eye bears a major respon¬ sibility for maintaining the sensory retina which lies above it. A prior art method has made possible routine visualization of the entire choroidal circulation, that is, all three vessel layers of the choroid can be visualized, superimposed one above the other. The innermost layer, the choriocapillaris, consti¬ tutes all of the nutritive vessels (i.e., where metabolic exchange with the retina takes place) for the choroidal circu¬ lation. The choriocapillaris layer occupies the plane immedi¬ ately adjacent to the sensory retina.
Although choroidal angiograms show all of the vessels of the choroid, information pertaining specifically to the chorio¬ capillaris is the most important, and there are conflicting views about the organization of the posterior pole choriocapil¬ laris, particularly concerning blood flow through it. The method of extracting information about the choriocapillaris from an indocyanine green (ICG) angiogram is therefore an important one to the clinician who is interested in evaluating the metabolic sufficiency and stability of the choroidal circulation.
Numerous investigators have used angiography and a variety of histological techniques to collect the current body of information about the choroidal circulation. Although the gross aspects of choroidal angioarchitecture and blood flow have been amply revealed by investigators' efforts, controver¬ sies still exist regarding regional differences in morphology. Additional controversies have also arisen regarding details of blood flow through this highly complicated vascular network. Of particular interest is blood flow through the chorio- capillaris, since, as discussed above, it is in this vascular layer that the nutritive function of the choroidal circulation takes place. Even though the state of the larger choroidal blood vessels must certainly influence choriocapillaris blood flow, ultimately it is a precise understanding of the chorio- capillaris blood flow itself that is fundamental to understand¬ ing the choroid's role in the pathophysiology of retinal disease.
High-speed indocyanine green (ICG) dye fluorescence angiography was developed to overcome the major problems encountered when attempting to visualize the rapid choroidal blood flow encountered in sodium fluorescein angiography. ICG angiography utilizes near-infrared wavelengths which penetrate the retinal pigment epithelium and choroidal pigment with relative ease. Whereas fluorescence from the choriocapillaris resulting from intravenously injected sodium fluorescein dye
(the other standard dye used in ocular angiography) appears to arise mainly from extravasated dye molecules or those adhering to the vessel walls, ICG fluorescence arises from dye molecules bound to blood protein in the moving blood volume. No doubt scanning laser ophthalmoscope fluorescein angiog¬ raphy (which can also utilize ICG dye) and the experimental technique of injecting fluorescein encapsulated in lipid vesicles eventually will produce additional information about choroidal blood flow; but with respect to clinical choroidal angiography, ICG angiography provides the best temporal and spatial resolution, making visualization of dye passage through the choroid possible under normal physiological conditions
(i.e., without having to artificially slow blood flow by such methods as raising intraocular pressure) .
When making intravenous dye injections, however, it is difficult to observe the choriocapillaris in individual ICG angiogram images due to the much higher levels of fluorescence arising from the large diameter underlying vessels. Due to this multi-layered organization of the choroidal vasculature, observation of the choriocapillaris with fluorescent dye angiography is best accomplished when a very small volume dye bolus having a sharply defined wavefront passes through. For example, following intra-carotid injection of a very small ICG dye bolus, ICG angiograms have been produced which clearly show the complete cycle of dye passage through an individual lobule under normal physiologic conditions. (Lobule is a term used to denote the three- to six-sided vascular units which form a mosaic pattern throughout the choriocapillaris. Each lobule consists of a cluster of narrow, tightly meshed capillaries which appear to radiate from a central focus at which a feeding arteriole enters at the posterior wall of the capillaries.)
Obviously, progression of a sharply defined wavefront is more easily tracked through the capillary network than an ill- defined one. Furthermore, if the bolus volume is small enough to essentially clear the underlying vascular layers by the time it enters the choriocapillaris, then images of the dye-filled capillaries will be of higher contrast than when significant fluorescence from beneath is simultaneously present.
Unfortunately, neither of the above conditions is readily produced by intravenous injection, even though passage of a dye bolus through the choroid can be optimized by appropriate injection technique. As a consequence, it is extremely diffi¬ cult to isolate choriocapillaris dye filling in raw ICG fluo¬ rescence angiograms even when they are recorded at high speed. Therefore, there is a need for a method that will make it possible to extract information about choriocapillaris filling from venous-injection ICG dye angiograms.
Despite their inability to provide complete information about the choriocapillaris, ICG fluorescence angiograms of the
choroidal circulation can delineate aberrant vascular struc¬ tures in the choroid which significantly diminish vision. Age- Related Macular Degeneration (ARMD) is the leading cause of significant visual impairment in the elderly. This disease is frequently characterized by development of choroidal neovas- cularization (CNV) membranes which invade the sub-retinal space, resulting in displacement of the sensory retina, and often blocking of the visual pathway as a result of subsequent hemorrhage. Treatment of ARMD is primarily by laser photocoagulation of the neovascular membrane. This treatment, however, is successful to the extent that the membrane can be accurately mapped; this is because such membranes are (by definition) in the macular area and often encroach on the fovea. Inappropri- ate application of photocoagulation can easily result in destruction of high acuity vision, and/or in accelerated growth of the CNV.
Diagnosis of and treatment of ARMD rely heavily upon interpretation of angiograms (both fluorescein and ICG) . Frequently, the morphology of CNV lesions is such that the membranes appear in fluorescein angiograms as little more than fuzzy blurs, if at all, especially when the membrane lies beneath a cirrus detachment. Moreover, today it is recognized that for a class of CNV, referred to as "occult-CNV", ICG angiograms provide necessary treatment data which sodium fluorescein angiograms cannot.
A further major difficulty in utilizing ICG angiograms when applying laser photocoagulation therapy is that the retinal vascular landmarks upon which the surgeon must depend when aiming the laser are often missing from the ICG angio¬ grams. The usual approach to resolving this problem is to make, during a separate setting, color photographs of the fundus and sodium fluorescein angiograms of the same eye of the patient; it is then necessary to attempt to superimpose the choroidal ICG angiogram and the retinal photograph or retinal fluorescein angiogram. This technique often fails due to the inability to precisely align the eye in exactly the same manner
during each of the two angiographic procedures. Nevertheless, very accurate alignment (within as little as 50 microns on the retina) is vital to safely apply laser photocoagulation near the fovea and, at the same time, assure no significant perma¬ nent damage to the fovea itself.
Therefore, there exists a need for new methods and devices to permit both better visualization of aberrant vascular struc¬ tures such as CNV and safer and more accurate laser photoco¬ agulation to rid the eye of such structures and improve vision.
SUMMARY OF THE INVENTION
The method of the invention is based on the premises that dye-filling of the choriocapillaris is more rapid —being pulsatile— than dye-filling of the underlying larger diameter vessels and that fluorescence from these two overlapping layers is additive. The premise regarding the velocity of blood in the choriocapillaris runs contrary to conventional wisdom regarding the relationship between blood velocities in parent and daughter vessels in most vascular beds.
In a nutshell, the invention consists of recognizing that pixel-by-pixel subtraction of an image from a succeeding image in an ICG angiographic sequence of images forms a resultant image sequence which shows fluorescence arising only from structures where the most rapid movement of blood occurs, i.e., in the choriocapillaris vessels.
This subtraction enhancement method of the invention makes it possible to extract information about choriocapillaris dye filling by taking advantage of the differences in large vessel and choriocapillaris blood flow rates which naturally exist. Instead of distinguishing choroidal layers by temporal sequence of dye bolus appearance, it is dye filling rates which serve to separate them.
Implementation of the invention depends only upon config¬ uring an existing fundus camera system to have sufficient temporal resolution and magnification of fundus structure. The described method was applied to high-speed ICG fluorescence
angiograms to emphasize information about choriocapillaris he odynamics.
In order to better visualize CNV and facilitate treatment of ARMD, however, the invention consists of a modified fundus camera with a polarizing filter in front of the excitation light source and an analyzing polarizer in front of the video camera. ICG dye fluorescence emanating from the fundus of the eye includes a significant component of polarized light, and rotation of the analyzer filter results in unwanted fluores- cence (i.e., that not associated with vascular structures, but rather associated with scattered light) being suppressed to the extent that the underlying CNV can be better visualized. This particular process affects the unprocessed, raw angiographic images in that it improves the signal-to-noise content of the individual angiographic images; subsequently, the subtracted raw images result in a clearer resultant image.
Once the aberrant vascular structure has been visualized and delineated by the polarization and subtraction methods but before laser photocoagulation therapy can begin, the surgeon must be assured that she can properly aim the laser. The invention further results from the usual practice of performing fluorescein angiography prior to performing ICG angiography and makes use of the fact that the fluorescein dye remains within the retinal vasculature for more than one hour. The invention utilizes an ICG fundus camera which has an integrating sphere coupled to light sources for excitation of both ICG and sodium fluorescein dye fluorescences and which uses a gatable charge-coupled device (CCD) video camera to capture the angiographic images. Light input to the integrat- ing sphere is via two fiber optic cables each connected to one of two light sources. One source is laser output at the wavelength needed to excite sodium fluorescein dye (480nm, i.e., a frequency-doubled Nd-Yag) ; it is also recognized that a shuttered, filtered incandescent light source can be used in place of a frequency-doubled laser. The other source is a diode laser output for excitation of ICG dye (805nm) .
As ICG dye transits through the choroidal circulation, the gated video camera records images of the ICG dye by causing the
805nm laser diode to fire in synchrony with the video camera. Appropriate programming of the camera and light sources are configured such that at regular intervals (e.g., every eighth image) the 480nm light source is fired and, simultaneously, an appropriate change is made in the barrier filter in front of the video camera.
To use the every-eighth-frame example, a barrier filter chain is implemented simply by placing a rotating disk contain¬ ing eight filters in front of the video camera. This filter wheel turns in synchrony with the camera firings such that every eighth frame corresponds to a positioning of the sodium fluorescein barrier filter in front of the camera. Because the sequence of angiograms is made at high speeds (approximately 15-30 images/second) , eye movements between successive images is insignificant, making precise registration of images trivi¬ al. Thus, the invention provides the ability to precisely superimpose the retinal vessel landmarks contained in sodium fluorescein angiograms on the delineated CNV lesions in the ICG angiograms, as needed by the surgeon to accurately focus a laser for treatment.
BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1, consisting of Figs, la and lb, illustrates an ICG fluorescence image of layers of ICG-stained blood to demon¬ strate fluorescence additivity and a graph produced from the image, respectively.
Fig. 2, consisting of Figs. 2a and 2b, illustrates sche¬ matically the brightness of fluorescent light emitted by two different blood vessels at times tj and t2, respectively.
Fig. 3, consisting of Figs. 3a, 3b, 3c and 3d, are, in 3a and 3b, ICG fluorescence images showing a 50 degree field of view centered on the macula of a right eye; the images were made 1/15 second apart. Fig. 3c is the result of subtracting the image of Fig. 3a from the image of Fig. 3b, and Fig. 3d is simply an enlargement of Fig. 3c. Fig. 4 illustrates a fundus camera system modified to provide the angiograms seen in Figs. 3a and 3b.
Fig. 5, consisting of Figs. 5a, 5b, 5c and 5d, illustrates four images of a left eye selected from a sequence of images produced by the subtraction method of the invention.
Fig. 6 illustrates a fundus camera system modified to suppress unwanted fluorescence.
Fig. 7 illustrates a fundus camera system modified to provide superimposed angiograms.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT Repeated real-time observations have shown that during ICG dye transit, after the large choroidal arteries fill, there is a rapidly pulsating faint and diffuse fluorescence superimposed over the steady fluorescence of the large vessels at the posterior pole. These pulsations appear to occur at a greater frequency than the heart rate, and they appear less obvious by the time the large choroidal veins are filled. Subsequent frame-by-frame analysis of the angiograms, however, indicate that the greater-than-heart-rate frequency is a perceptual phenomenon resulting from the out-of-phase pulsatile filling of individual lobules, all at near-heart-rate frequency. Unfortunately, not enough is known yet about details of choriocapillaris hemodynamics to account with certainty for the observed more rapid fluorescence intensity changes in the choriocapillaris than in the larger underlying vessels, but the most likely reason is that choriocapillaris blood flow velocity is greater than that through the underlying choroidal vessels. The invention is based on the premises that the fluorescence intensities of ICG-filled choriocapillaris and underlying vessels are additive and that there are detectable differences in the rates of change of fluorescence intensities emanating from the choriocapillaries and the underlying choroidal vessels as they fill with dye.
Although the average cross-sectional diameter of the choriocapillaris is much smaller than that of the underlying arterial and venous vessels which feed and drain them, it
appears that fluorescence from the two vascular layers is additive. ICG fluorescence additivity was demonstrated by creating a stair-step wedge of overlapping thin layers of heparinized blood containing ICG dye (0.03 mg/ml) ; each step was formed by a thin layer of the blood sandwiched between two microscope slide coverglasses.
Fig. la shows an ICG fluorescence image of the stair steps. The horizontal white line through the center of the image indicates the path along which image pixel brightness (i.e., grey level) was measured to produce the graph in Fig. lb, demonstrating stepwise increase in fluorescence as the number of overlapping blood layers increased.
The greater rate of change in dye fluorescence intensity in choriocapillaries than in the larger underlying vessels is shown schematically in Figs. 2a and 2b. In Fig. 2a, the brightness of a large diameter vessel and an overlying chorio¬ capillaris vessel (both in cross-section) are indicated as vectors, IA and Ic, respectively. The fluorescent light emit¬ ted by both is detected at time tj by a light sensor, S. In Fig. 2b, the status of the same two vessels and sensor is shown at later time t2, where ΔlA and Δlc are respectively the incre¬ mental increases in brightness of the two vessels. Therefore, the total brightness detected by the sensor at tx is:
Sti = IA + Ic
At time t2, the total brightness detected is:
St2 = IA + Ic + ΔIA + ΔIC
The change in total detected brightness which occurred between
ΔS = St2 - Stl - ΔIA + ΔIC
10 But since ΔlA « Δlc, ΔS = Δlc.
In other words, the small change in the combined bright¬ ness of the overlapping capillary and large vessel which occurs during a short time interval is virtually all attributable to the choriocapillaris vessel. This phenomenon can be demon¬ strated by the method of the invention, i.e., by subtracting, pixel for pixel, an image in a high-speed ICG fluorescence angiogram sequence from a succeeding image, as demonstrated in Figs. 3a-d. Figs. 3a and 3b are angiographic images made 1/15 second apart. Fig. 3c is the result of subtracting those two images, and Fig. 3d is simply an enlargement of Fig. 3c.
Note that in the resultant image (Fig. 3c or 3d) lobular structures are seen which were not apparent in either of the original images (Fig. 3a or 3b) . Also, instead of the dye- filled retinal arteries seen in the original images, only a dye wavefront representing the movement of additional dye into the retinal arteries near the disc is seen in the resultant image. Of course, the more spatially well defined the dye bolus, the more dramatic is the effect of the invention. Not all intrave- nously injected dye boluses produce as dramatic results as were achieved in this example, but in each case there is enhancement of the choriocapillaris component of fluorescence. Note, the subtraction method of the invention is intended to operate by subtracting the image from any succeeding image. To test the method of the invention, five normal rhesus monkeys between two and three years of age were used. For each observation a monkey was immobilized by intramuscular injection of ketamine hydrochloride (10 to 15 mg/kg) , intubated, and then maintained lightly anesthetized with halothane; mydriasis was induced by topical application of 1% tropicmide. Small boluses (about 0.05 ml) of ICG dye (12.5 mg/ml) were injected through a catheter inserted in the greater saphenous vein and immediately followed by a 2.0 ml saline flush. Passage of dye through the choroidal vasculature was detected using a modified Zeiss fundus camera and directly digitally recorded by PC-based video frame-grabbers. At least three angiographic studies of the same eye were performed on different days for each monkey.
In the above test, as shown in Fig. 4, the usual fundus camera 10 was modified by replacing the xenon flash tube light source with an 805 nm wavelength laser diode 12 coupled to the fundus camera's illumination optics 14 via a small integrating sphere 16 whose exit port was located at the position normally occupied by the flash tube arc. The fundus camera's usual means for receiving images, i.e., the photographic film camera, was replaced with an infrared sensitive vidicon tube (model 4532URI Ultracon, Burle Industries) 18 (a charge-coupled device could be used instead of the vidicon tube) , in front of which an 807 nm wavelength cut-on filter 20 was placed to exclude the excitation laser light while admitting ICG dye fluorescence light. Choroidal dye transit was recorded in thirty-two consecutive video angiographic images at a rate of 30 or 15 frames per second by two digital frame grabbers (model 2861-60, Data Translation) (not shown) installed in a personal computer (Compaq, model 386/25e) (not shown) .
Fig. 5 summarizes the angiographic findings obtained in the above test by applying the image subtraction method of the invention. In this example case, each image in a 15 frames/second ICG angiographic sequence was subtracted from the image immediately following it; the images in Fig. 5 were selected from the resulting sequence of subtracted images.
Dye first enters the macular area of the choriocapillaris which lies temporal to and above the points at which the short posterior ciliary arteries enter the eye (Fig. 5a) . A lobular pattern can be seen in the center of the angiogram, particular¬ ly just nasal to the center; here a cluster of unfilled lobules is shown (arrows). 0.133 seconds later (Fig. 5b) the entire central area is completely filled, although two smaller clus¬ ters of late-filling lobules may be seen superior to the center (arrows) . Choriocapillaris filling progresses almost radially from the macular region. By close inspection of this image, faint loss of fluorescence around lobules can be seen; these likely correspond to choriocapillaris drainage channels.
Fig. 5c is 0.200 seconds later than Fig. 5b. It indicates that the radially oriented wave of choriocapillaris dye filling has been completed, and dye distribution at the posterior pole region appears fairly uniform. This image indicates that the first wave of dye filling is complete within the center of the macular region, as indicated by the appearance of relatively hypo-fluorescent areas which were hyper-fluorescent in Fig. 5a. In Fig. 5d, 0.133 seconds later, it appears that the first wavefront of dye filling has reached the peripheral region; at this stage, Fig. 5d is nearly a complete reverse contrast image of Fig. 5a.
The wavefront of dye filling traveled radially from the macular region to the periphery of the 30 degree field of view in approximately 0.466 seconds. This overall filling pattern was present in each eye observed, and details of the filling patterns were remarkably consistent from observation to obser¬ vation for each subject eye.
ICG fluorescence angiography gradually is being used more frequently by both researchers and clinicians to investigate the choroidal circulation. Clearly, as such new tools are applied in a variety of new ways to studying the choroid, old concepts about it and its physiology will be revisited, and some will change or give way to entirely new concepts. Fortu¬ nately, some approaches to analyzing choroidal angiograms like the subtraction method of the invention described above may be applied both in animal and in human clinical research with complete safety, perhaps hastening a better understanding of choroidal blood flow in health and disease.
ICG fluorescence angiography is used in the diagnosis and treatment of ARMD; however, as noted above, the difficulty arises in attempting to accurately map choroidal neovasculari- zation (CNV) . The invention lies in recognizing that fluores¬ cence arising from a dye molecule contains information about the processes that take place within the molecule during the time between excitation and emission of light by the molecule.
Moreover, fluorescence of molecules can be affected by the characteristics of the substances to which the molecule is bound and by the character of the binding which has taken place.
For example, in the case of ICG dye in the vasculature of an eye containing CNV, the dye may bind with greater affinity to neovascular endothelium than to established endotheliu . In such a case, fluorescence arising from those bound dye mole¬ cules may be substantially different from fluorescence associ¬ ated with ICG dye molecules which may be bound to other types of protein in the cirrus fluid or from ICG fluorescent light simply scattered by the presence of protein molecules within the cirrus fluid. In either event, ellipsometry is an appro¬ priate tool for improving the visualization of CNV.
The invention then, as shown in Fig. 6, is a modified fundus camera 22 with a polarizing filter 24 in front of the excitation light source 26 and an analyzing polarizer 28 in front of the video camera 30. ICG dye produces a high degree of polarized ability, and rotation of the analyzer filter results in the fluorescence from the cirrus fluid being sup¬ pressed to the extent that the underlying CNV can be better visualized. This particular process affects the unprocessed, raw angiographic images in that it improves the signal-to-noise content of the individual angiographic images; subsequently, the subtracted raw images result in a clearer resultant image.
Once an aberrant vascular structure such as CNV is clearly delineated, it can be treated using laser photocoagulation therapy; however, as noted above, aiming the laser properly requires superimposing an ICG angiogram and a retinal photo¬ graph or retinal fluorescein angiogram. The invention results from the usual practice of performing fluorescein angiography prior to performing ICG angiography making use of the fact that the fluorescein dye remains within the retinal vasculature for quite long periods of time (more than one hour) . Therefore, if one configures an ICG fundus camera in such a way that during
the course of obtaining ICG angiograms, a fluorescein angiogram can be obtained (within fractions of a second of obtaining a previous and succeeding ICG angiogram) , no significant movement of the eye can take place. This means that the intervening fluorescein angiogram would, by definition, precisely register with the ICG angiograms.
As shown in Fig. 7, the invention utilizes an ICG fundus camera 32 which has an integrating sphere 34 coupled to light sources for excitation of ICG dye fluorescence and which uses, as an image receiving means, a gatable video camera 36 (prefer¬ ably CCD) to capture the angiographic images. Light input to the integrating sphere is via two fiber optic cables 38, 40, each connected to one of two light sources 42, 44; one source 42 output is at the wavelength needed to excite sodium fluores- cein dye (480nm) and the other source 44 output for excitation of ICG dye (805nm) .
As ICG dye transits through the choroidal circulation, the gated video camera 36 records images of the ICG dye by causing the 805nm laser source 44 to fire in synchrony with the video camera 36. Appropriate programming of the camera and light sources are configured such that at regular intervals (e.g., every eighth image) the 480nm source 42 is fired, and simulta¬ neously an appropriate change is made in the barrier filter 46 in front of the video camera. To use the every-eighth frame example, the barrier filter chain is implemented simply by placing a rotating disk contain¬ ing eight filters in front of the video camera. This filter wheel turns in synchrony with the camera firings such that every eighth frame corresponds to a positioning of the fluores- cein barrier filter in front of the camera. Thus, the inven¬ tion provides the ability to precisely superimpose angiograms needed by the surgeon in order to accurately aim a laser photocoagulation beam.