2.1.

Regimes of Operation

The system can operate in different modes. In the B-scan mode, only one galvomirror of the galvanometer scanning pair is driven with a ramp at 700 Hz, and the translation stage is moved for the depth range required in 0.5 s. In this case, an OCT B-scan image is produced in either the x,z -plane or the y,z -plane.

In the C-scan mode, one galvoscanner is driven with a ramp at 700 Hz and the other with a ramp at 2 Hz. In this way, C-scan images in the x,y -plane are generated in both channels. The depth at which the OCT C-scan image is collected is changed by moving the translation stage in the reference arm of the interferometer in Fig. 1.

The frame grabber has two inputs, each equipped with an 8 bit analog-to-digital (A/D) converter, and displays a pair of two images, each with 480 lateral and 350 vertical pixels irrespective of the regime of operation, B-scan or C-scan. The images are pixel-to-pixel correspondent and such pixels are constructed simultaneously in the two images, first along lines in the rasters and then from the end of the line to the beginning of the next consecutive line. Therefore the two images are acquired and constructed simultaneously.

In the B-scan regime, the frame grabber operates under the control of the generator driving the galvanometer scanner and the translation stage. Each line in both images is triggered by the generator driving the galvanometer scanner (700 Hz), while the generation of the frame of the pair of images is triggered by the translation stage (every 0.5 s). The horizontal size of the OCT is determined by the magnitude of the signal applied to the galvanometer scanner, while the vertical size corresponds to depth. The image in the confocal channel consists of repeated distributions of the intensity of the signal collected. The horizontal size is the same as that of the OCT image, while the vertical size represents time.

In the C-scan regime, the frame grabber operates under the control of the generators driving the two galvanometer scanners. Each line in both images is triggered by the faster generator (700 Hz), while the generation of the frame of the pair of images is triggered by the slower generator (2 Hz). The horizontal size of the OCT and confocal images is determined by the magnitude of the signal applied to the fast galvanometer scanner, while the vertical size in both images is determined by the magnitude of the signal applied to the slow galvanometer scanner. Operating at two frames a second, this means that the two images are constructed from the top left corner pixel, at coordinates (1,1), then line by line till the end of a frame period of 0.5 s, when the last pixel (480, 350) in the bottom right corner of each image is displayed.

In comparison with previous reports,¹⁴ we have improved the interface optics to expand the field of images collected from the retina and include both the fovea and the optic nerve in a single frame. Such images from the right eye of one of the authors (AP), illustrating the two regimes of operation are shown in Fig. 2. Each pair of images was obtained in 0.5 s. The pair of images in Figs. 2(a) and 2(b) was obtained in the B-scan regime, while the pair of images in Figs. 2(c) and 2(d) was obtained in the C-scan regime. The images in Figs. 2(a) and 2(c) are delivered by the confocal channel while the images in Figs. 2(b) and 2(d) are delivered by the OCT channel. The signal in the confocal channel is determined by the reflectivity of the target, while in the OCT channel it is determined by the square root of the same reflectivity. In other words, a strong change in the index of refraction from a scattering point to the surrounding medium is indicated by a signal of high intensity in both channels.

Figure 2

Images of a normal eye generated with the stand-alone OCT-confocal system showing the fovea and the optic nerve. Transverse pixel size, <20 μm; depth pixel size, <13 μm in tissue. Lateral size in all images, 35 deg (horizontal). (a) and (b) B-scan mode: Only the horizontal transverse scanner is driven; (a) is a repetition of linear en face scan profiles in time, at different moments while the depth in the range of 1 mm in the OCT channel is changed to produce the longitudinal OCT in (b). The 1-mm depth is along the vertical axis in (b) only, measured in air. The image features details of the optic nerve, foveal depression, and inner retinal layers, such as the retinal nerve fiber layer and the retinal pigment epithelim. (c) and (d) C-scan mode: Both the horizontal and vertical transverse scanners are driven; (c) is the confocal image in the pair and (d) is a deep C-scan OCT at the level of the choroid, demonstrating its characteristic vascular lobules. Vertical size, 15 deg.

Achieving an image size of a 35-deg angle was made possible by further refinement of the scanning elements and processing electronics to cope with the subsequent increase in the bandwidth. The image bandwidth is proportional to the scanning rate and image size, and in this case components up to 300 kHz are generated. We have, however, slightly compromised the transverse size resolution and reduced the electronic bandwidth to 200 kHz to achieve a signal-to-noise ratio before the frame grabber of more than 40 dB, a value close to the dynamic range of the 8-bit A/D converter used in the frame grabber. In evaluating the signal-to-noise ratio, we used the signal reflected from the retinal pigment epithelium (RPE).

However, in interpreting the pathology cases that follow, we present smaller size images (15 and 20 deg) for better display of tiny features. When collecting smaller sized images, we maintained the same electronic bandwidth of 200 kHz and the same frame rate, 2 Hz.

In the images presented in this paper, no other phase modulation was employed apart from that introduced by the X-galvanometer scanner. We previously demonstrated the role played by the image size¹² in balancing the effects of an external phase modulator and of the modulation produced by the transversal scanner. For instance, when the image size was six times smaller than that of the images in Fig. 2, it was advantageous to employ a phase modulator working at f=30 kHz.

2.2.

Resolutions

Because the focus is not changed when the path imbalance in the OCT is altered, the confocal image does not change with depth z. Therefore all the lines in Fig. 2(a) are essentially the same; each line is the linear variation of the intensity received along the X -axis at different moments of time during the frame time interval. Similarly, the image in Fig. 2(c), which is the (x,y) map of the intensity, is the same in all the pairs of images collected at different positions of the translation stage.

Using a plane mirror in the object arm, the FWHM of the depth sampling interval in both channels was determined. To determine the depth resolution of the OCT channel, a very small amplitude was applied to the galvanometer scanner operating at 700 Hz to create a path modulation larger than λ/2, while no voltage was applied to the other galvanometer scanner. Using the translation stage, shifted either side of the maximum of interference until the intensity fell to half, a depth resolution in air of 18 μm was obtained for the OCT channel. The depth sampling interval in tissue can be obtained by dividing this value by the index of refraction (this varies, depending on the tissue; however, an average value of 1.36 can be used¹⁵), leading to about 13 μm. Obviously, this is the ultimate resolution achievable for stationary tissue only. The interpretation of the images in Figs. 2(b) and 2(d) should consider that the two frame/s rate is relatively low compared with the equivalent axial speed resulting from microsaccades, blood pulsations, and head movements. Therefore the images will not be artifact free in this respect.

To determine the FWHM of the depth profile of the confocal channel, the voltages on the galvanometer scanners were brought to zero and the mirror moved through focus until half of the value of the signal obtained in the focus was obtained, which led to an FWHM of 5.3 mm in air.¹⁴ This value was determined by the pinhole size (30 μm) and the focusing lens (1.5-cm focal length microscope objective) in the confocal receiver, which have been adjusted as a tradeoff between the image strength and elimination of stray reflections from the interface optics and the cornea. The choice of pinhole size is related to the beamsplitter ratio;¹⁴ the smaller the percentage diverted from the backscattered signal to the confocal receiver, the larger the pinhole size should be. Since the depth-resolved information is provided by the OCT images, the depth resolution of the confocal channel is less important as long as the stray reflections are sufficiently attenuated. The role of the confocal channel is to provide lateral and contextual information, owing to its better transverse resolution, and in this way the two imaging systems complement each other.

Because the confocal receiver is far from ideal,¹⁴ the transverse resolution in the confocal channel is practically that of a conventional microscope. The beam diameter output from the instrument to the eye is 3 mm. This should determine a lateral resolution better than 5 μm. It is known, however, that owing to eye aberrations¹⁶ and scattering, an ideal confocal channel for a 3-mm aperture is not expected to achieve better than a 10 to 15-μm transverse resolution from the back of the eye.¹⁷

Owing to the way the carrier is created in the OCT channel^{10 11 12} and the speckle effects, it was expected that the transverse resolution in the OCT channel would be worse than that of the confocal channel. However, using a plane target of holes and lines at the back of a lens of 2 cm focal length, the transverse resolution was found to be better than 20 μm in both channels.

3.1.

En Face Scanning Allows High Transversal Resolution

As a result of transverse scanning, the B-scan OCT image in our system is continuous along the line in the raster, as opposed to B-scan OCT images generated using fast axial scanning, where the lateral scanning is discrete. This improves the quality of the images, as demonstrated later. Although our OCT system has only a 12-μm depth resolution in tissue, the B-scan image in Fig. 2(b) looks similar to the B-scan image generated by the group at Massachusetts Institute of Technology¹⁸ using longitudinal OCT and a much higher depth resolution, of 3 μm in tissue. Any connectivity or association between scatterers in the transverse section is better conserved when performing en face scanning. This ultimately allows visualization of small protuberances on the retina, as demonstrated in this paper with several cases of pathology.

3.2.

Synergy between the Channels

In terms of data acquisition, the confocal image adds further versatility. The design ensures a strict pixel-to-pixel correspondence between the two C-scan images (OCT and confocal). This helps in two respects: for small movements, the confocal image can be used to track the eye movements between frames and for subsequent transversal alignment of the OCT image stacks; for large movements and blinks, the confocal image gives a clear indication of the OCT frames that need to be eliminated from the collected stack. As a reference for the aligning procedure, the first artifact-free confocal image in the set is used. For example in Fig. 2(a), movements of the eye are indicated by lateral shifts of the confocal traces. Each horizontal line in the confocal image corresponds to a depth position. The relative eye movement, proven by the slight deviation of shadows to the right, can easily be utilized to correct the lateral shift of the lines in the B-scan OCT image in Fig. 2(b).

3.3.

3-D Imaging

3-D imaging of the retina is already common with CSLO technology.¹⁹ The use of en face sections collected from different depths²⁰ to construct a 3-D profile of the retina is already accepted and understood by ophthalmologists. The stand-alone OCT-confocal system can proceed in the same way, however, with en face slices as thin as allowed by the OCT technology. To collect the reflectivity distribution from the volume of the retina, the stand-alone OCT-confocal system is operated in the C-scan mode, collecting en face images at different depths. Ideally, the depth interval between successive frames should be much smaller than the system resolution in depth, and the depth change should be applied only after the entire C-scan image is collected. However, in practice, to speed up the acquisition, the translation stage is moved continuously and the depth interval is set slightly larger than the depth resolution.²¹ For instance, using 20 μm between frames, 60 frames of image pairs from a volume in depth of 1.2 mm in air (which is sufficient to cover the volume of the retina around the optic nerve) can be acquired in 30 s when operating at a 2-Hz frame rate. After acquisition, the images can be aligned transversally using the first confocal image (or the first consistent confocal image) and then the stack of OCT images or the stack of the pairs of OCT and confocal images can be used to construct a 3-D profile of the volume of the retina.^{21 22} It is much easier to align the pairs of images in this way than aligning C-scan OCT images that have different features for different depths.

3.4.

Topography

The topography of the fundus is difficult to obtain using A-scan images. A procedure using A-scans is cumbersome because it requires interpolation in the en face plane.²³ Owing to a wide variety of features being available to use for interpolation, with many of these being of high contrast, it is likely to be more accurate to construct the topography using en face scanning. The topography in itself outputs an en face image, which makes the en face OCT more suited for signal processing. Such a procedure using collected en face images was described in a previous report.²⁴ Based on en face OCT, obtaining a topography should be similar to the method using a CSLO.²⁵

3.5.

New Challenge

New imaging technology not only brings new information to the clinician, but with it the requirement of interpretation. En face OCT is no exception in this respect. The higher the depth resolution of the OCT system, the more fragmented the en face OCT image looks.²⁶ Since the imaging proceeds at a few frames a second, the inherent eye movements may result in significant changes in the size of fragments sampled from the tissue. The fragmentation is especially visible when imaging very inclined tissue. The en face OCT image in Fig. 2(d) shows the challenges in interpreting and using these images. First, the en face OCT image looks fragmented, and on its own, such an image cannot be interpreted. Second, variations in tissue inclination with respect to the coherence wave surface alter the sampling of structures within the depth of the retina,²⁷ producing novel slice orientations that are often challenging to interpret.

The bright patches in the OCT image represent the intersection of the surface of the optical path difference (OPD)=0 with the tissue. Owing to the particular way the retina is scanned, with the fan of rays converging on the eye’s pupil, the surface of OPD=0 is an arc circle with the center in the eye’s pupil. When we explore the depth, we change the radius of the arc. If the arc has a small radius, it may just only intersect the top of the optic nerve, with the rest of the arc in the aqueous. The radius of the arc is changed by changing the length of one of the arms of the interferometer in the OCT channel to explore the retina up to the RPE and choroid. The orientation of the retina tissue at the back of the eye is not planar, and this complicates the interpretation of the image even further.

Another issue is that despite scanning images en face, the images may display the structure in depth, as in any longitudinal OCT image. These two effects—fragmentation and depth structure—displayed in the C-scan images are present in a CSLO with high depth resolution as well; however they occur at a scale where they are regularly discarded. In a CSLO, the images do not look fragmented and the depth structure is barely visible, owing to the coarse depth resolution (0.3 mm), which is comparable to the retina’s thickness. Going in and out of focus results in a smooth transition from dark to bright of areas in the image. Both problems mentioned earlier are brought about by the high depth resolution of OCT. We address the fragmentation problem by providing the confocal image, which guides the user, and by collecting many en face images at different depths and subsequently building the 3-D profile. The other problem—that the en face image may also display the depth structure—requires further development of the interpretation process.