1. Introduction

The retina is responsible for effective capture of photons and several preliminary stages of visual information processing. Advanced investigation of information processing in the retina is important for better understanding of the nature of vision, and thus allows better prevention of vision losses. Single and multiple-channel electrophysiological recording provide valuable information for study of retinal information processing [1]. However, spatial resolution of conventional electrophysiological measurement is limited. Low spatial resolution limits the capability of electrophysiological measurement for accurate assessment of localized retinal activity. Given the delicate, complex layered structure of the retina [2], better study of retinal information processing requires the capability to monitor dynamic activity of large populations of retinal neurons simultaneously, with high spatial resolution in three dimensions.

Stimulus-evoked intrinsic optical signals (IOSs) have been detected in photoreceptor outer segments [3], isolated retinas [4, 5], and eyecup slices [6]. Fast IOSs closely associated with action potentials and postsynaptic potentials have been also observed in other neural tissue, such as dissected axon fibers [7, 8]. However, practical applications of fast IOSs for functional imaging of neural systems have been challenged by the small magnitude and limited signal-to-noise ratio (SNR) of IOSs [9]. Our recent investigations indicate that the sensitivity and SNR of intrinsic optical signal recording could be substantially improved through optimized near infrared (NIR) light illumination and improved spatial resolution, allowing high performance optical imaging of retinal activity with single pass measurements [10]. Dark-field and polarization-sensitive imaging could further improve the sensitivity of intrinsic optical signal imaging [11, 12]. NIR light imaging of fast IOSs promises simultaneous monitoring of many retinal neurons, with high spatial resolution. However, intrinsic optical signal imaging of retinal information processing is still challenged by unclear sources and mechanisms of fast IOSs. Both positive and negative intrinsic optical signals have been reported [10, 13–16]. Better understanding of the sources and biophysical mechanisms of fast IOSs are important for pursuing intrinsic optical signal imaging of dynamic information processing in the retina.

While isolated retinas provide a simple preparation for study of early visual information processing; in vivo recording of fast IOSs may provide new method for noninvasive diagnosis of retinal function. By using a corneal electrode to record light-evoked electrophysiological responses of the retina, conventional full-field electroretinogram (ERG) [17] and recently emerging multifocal ERG [18, 19] measurement can be used to assess the physiological health of photoreceptors and inner neurons. ERG measurement of stimulus-evoked retinal activation is sensitive and practical for clinic application. However, relatively low spatial resolution and lack of direct morphological information make it difficult to provide accurate identification of localized retinal dysfunction. The ophthalmological community is actively making efforts to pursue in vivo detection of IOSs associated with retinal activation. Several imaging techniques have been demonstrated for intrinsic optical signal imaging of retinal activation. Conventional fundus cameras have been modified to detect IOSs in anesthetized animals, such as cats [20] and macaques [21], and awake humans [22]. Both time-domain and frequency-domain optical coherence tomography (OCT) imagers have been used for depth-resolved recording of IOSs in isolated frog retina [15], isolated rabbit retina [13], and living rat eye [23]. Both flood-illumination [14] and laser scanning [24] adaptive optics ophthalmoscopes have been validated for in vivo imaging of IOSs in living human. Using a flood-illumination adaptive optics imager, transient IOSs in individual cones has been detected [14].

Although in vivo recordings of IOSs have been successfully demonstrated, practical application of intrinsic optical signal imaging for retinal diagnosis is still challenged by the complexity and inconsistency, in terms of time courses and signal polarities, of reported IOSs [24]. In general, both stimulus-evoked retinal neural activity and corresponding hemodynamic and metabolic changes may produce IOSs. While hemodynamic and metabolic changes associated IOSs can provide important information for healthy assessment of visual system, they are relatively slow and cannot directly track fast neural activities in the retina. Fast IOSs that have time courses comparable to ERG responses are desirable for reliable assessment of photoreceptors and inner neurons directly. In principle, high-speed imaging systems, such as adaptive optics [14] and OCT [23] imagers, can provide temporal resolution required for imaging fast IOSs associated with ERG responses, such as a- and b-waves. ERG a- and b-waves are most often measured to assess physiological health of photoreceptors and inner neurons, respectively. However, in vivo imaging of fast IOSs is still difficult [24]. Better understanding of the sources and biophysical mechanisms of fast IOSs may provide insight to optimize instrument designs and test protocols to pursue reliable measurement of fast IOSs, which have time courses comparable to ERG responses.

Isolated retinas have been extensively used for study of early visual information processing in the retina, and may also act as a translation sample to simplify experimental procedures and controls for further development of retinal imaging equipment. Without complicated involvement of stimulus-evoked hemodynamic changes, such as dynamic blood flow and oxygen changes, isolated retinas provide a simple preparation for better characterization of fast IOSs. Using a flood-illumination microscope equipped with high-speed digital camera, we have observed both positive and negative IOSs in isolated amphibian retinas [10, 11]. Our recent investigation suggested that the negative and positive IOSs result primarily from photoreceptors and inner neurons, respectively [16, 25]. Using a home-built functional OCT, we have demonstrated depth-resolved recording of IOSs in frog retina, but fast enface OCT imaging of individual retinal layers was not practical because of the time-consuming scanning requirement of our OCT system [15]. Depth-resolved enface imaging is important for better visualization of neural activity at individual functional layers of the retina.

In this article, we describe depth-resolved enface imaging of fast IOSs in isolated frog retina. A flood-illumination microscope, equipped with high numeric aperture (NA) objective, provided a practical strategy to pursue fast enface imaging of IOSs. By adjusting the NIR light imaging depth relative to retinal surface, we captured intrinsic optical signal patterns from the photoreceptor and inner retinal layers. Both negative and positive IOSs were observed; and transient ON and OFF responses were observed associated with the step light stimulus. High temporal (ms) resolution NIR light images disclosed fast IOSs within 5 ms after the stimulus onset. Our experimental results consistently support the hypothesis that positive IOSs result primarily from the inner neurons; while negative IOSs mainly relate to photoreceptor response. However, dynamic phototransduction procedures may partially contribute to the positive signals observed at the photoreceptor layer. The optical wave guiding property and center-surround antagonism of the retina may also affect the production and redistribution of the IOSs.

2. Methods

2.1. Preparation of isolated retinas

Isolated retinas of Leopard frogs (Rana Pipiens) were used for the experiments. All experiments were performed following protocols approved by the Institutional Animal Care and Use Committee of the University of Alabama at Birmingham. 20–30 minutes dark-adaptation was typically implemented before the dissection of the retina. During the experiment, the frog was rapidly euthanized by decapitation and double pithing before the eyes were removed. The procedure was conducted in a dark room with dim red illumination. The dissection of the retina was performed in Ringer’s solution containing (in mM) [26]: 110 NaCl, 2.5 KCl, 1.6 MgCl₂, 1.0 CaCl₂, 22 NaHCO₃, and 10 D-glucose. After removing the intact eye, we hemisected the globe below the equator with fine scissors to remove the lens and anterior structures before removing the retina without retinal pigment epithelium (RPE). The isolated retina was sliced radially, thus allowing it to lie flat in the recording chamber. Finally, the retina was transferred into the recording chamber for optical imaging. During the recording, the retina was immersed in a chamber filled with Ringer’s solution. The retina was pressed to the bottom of the sample chamber with moderate pressure using a micromesh sheet [15].

2.2. Experimental setup

A NIR light flood-illumination microscope was constructed to measure dynamic IOSs associated with retinal activation (Fig. 1). Instead of a 4X (NA=0.13) objective used for our previous studies [10, 11, 16, 25], here we use a 10X (NA=0.3) objective to achieve the required axial resolution. In principle, the axial resolution, i.e., the depth of field, of the 10X (NA=0.3) objective is about 10 µm, and thus allows depth-resolved enface imaging of IOSs from different retinal layers. During the experiment, the retina was illuminated continuously with NIR light for recording transient IOSs, and a visible light flash was used for retinal stimulation.

 

Fig. 1. Schematic diagram of the experimental setup. During the measurement, isolated frog retina was illuminated continuously by the NIR light for recording of stimulus-evoked IOSs. The visible light stimulator was used to produce a visible light flash for retinal stimulation. Concurrent ERG measurement was conducted to record electrophysiological responses associated with retinal activation. At the dichroic mirror (DM), visible stimulus light was reflected and NIR recording light was passed through. The eyepiece camera was used to adjust visible light stimulus aperture at the retina. In order to ensure light efficiency for intrinsic optical signal imaging, the beam splitter (BS) was removed from the optical path after we adjusted the visible light stimulator. The NIR filter was used to block visible stimulus light, and allow the NIR probe light to reach the CCD/CMOS camera.

Download Full Size | PDF

The NIR light was produced by using a halogen lamp with a band-pass filter (wavelength band: 800–1000 nm). The visible light stimulator was a fiber-coupled white light emitting diode (LED) with central wavelength at 550-nm (wavelength band: 450–650nm). At the retina, the diameter of the stimulus aperture was ~60 µm. By adjusting the NIR light imaging depth relative to retinal surface, we were able to selectively record images from photoreceptor, inner nuclear (i.e., middle), and ganglion layers of the retina. The images shown in Figs. 2–5 were recorded with a 14 bit CCD camera (PCO1600, PCO AG, Kelheim, Germany). The CCD camera has 2 GB built-in random-access-memory (RAM) for fast image recording with a transfer speed of 80 MB/s. Although the camera has relatively high resolution (1600×1200 pixels), we often used it at a lower resolution employing pixel binning for high frame rate recording. The CCD images presented in this article were recorded with a speed of 80 frames/s, and each frame consists of 320×240 pixels. We also used a 10 bit CMOS camera (PCO1200, PCO AG, Kelheim, Germany) to characterize the ON and OFF responses of the IOSs associated with the step light stimulus. The CMOS camera also has 2 GB built-in RAM for fast image recording with a transfer speed of 820 MB/s. Because of the ultrafast transfer speed of the CMOS camera, intrinsic optical signal images could be collected with high frame speed with sufficient exposure time to ensure image quality. The CMOS images presented in this article were recorded with a speed of 1000 frames/s, and each frame consists of 160×120 pixels. The CMOS camera allowed us to characterize fast intrinsic optical responses with millisecond temporal resolution and micrometer spatial resolution.

2.3. Intrinsic optical signal imaging and data processing

The retina consists of multiple functional layers [2]. For an isolated frog retina, the whole thickness is 150–200 µm. By adjusting the axial (depth) location of the NIR imaging plane, intrinsic optical signal images were recorded from the photoreceptor, inner nuclear (middle), and ganglion layers. Concurrent ERG measurement was conducted to monitor the electrophysiological response associated with retinal activation. For each experimental measurement shown in Figs. 2–5, the frog retina was activated by a single light flash (125 ms). The stimulus intensity was equivalent to ~4.0×10⁴ [550-nm photons]/ms·µm^-2. This stimulus strength was selected to evoke both negative and positive IOSs in isolated frog retina [16]. Under this experimental condition, IOSs were robust for at least 1 hour, with a stimulus interval of 2 minutes. In other words, at least 30 imaging passes could be implemented with an isolated frog retina without RPE. In order to ensure the reproducibility of the IOSs, two imaging sequences were typically recorded from each retinal layer sequentially, with identical stimulus and recording parameters. In order to assess dynamic optical changes associated with retinal ON and OFF responses, we used prolonged (500 ms) step stimulus for the experiment shown in Fig. 6. The three imaging sequences shown in Fig. 6 were recorded from the photoreceptor, inner nuclear, and ganglion layers sequentially, with a time interval of 5 minutes.

The intrinsic optical images presented in this article show dynamic optical changes with unit of ΔI/I, where ΔI is dynamic optical response associated with retinal activation and I is background light intensity. The intrinsic optical signal images were reconstructed from the CCD/CMOS images as follows: 1) The pre-stimulus baseline images were averaged, pixel by pixel, and were taken as the background light intensity I of each pixel; 2) The background light intensity I was subtracted from each recorded frame, pixel by pixel, to get the dynamic change ΔI of each pixel of the images. 3) Image sequence of ΔI/I was reconstructed to show the stimulus evoked transient intrinsic optical changes in the retina.

3. Results

3.1. CCD imaging of IOSs evoked by single flash stimulus

Using the CCD camera, depth-resolved NIR light enface imaging was implemented to characterize IOSs at photoreceptor, inner nuclear, and ganglion layers of the retina. Figures 2–4 represent typical IOSs recorded from the photoreceptor, inner nuclear, and ganglion layers. These intrinsic optical signal images were sequentially recorded from the same lateral location of a frog retina, with consistent stimulus (duration: 125 ms; intensity: 4.0×10⁴ [550-nm photons]/ms·µm^-2). We recorded two imaging sequences from each layer to ensure the reproducibility of the IOSs. Our experimental results indicate that transient IOSs can be recorded from stimulus-activated retina, with single pass measurements. Both positive and negative IOSs were observed during retinal activation.

 

Fig. 2. (a) Representative CCD image sequence of photoreceptor layer, without differential processing. The white spot in the second frame shows the visible stimulus pattern. At the photoreceptor layer, the diameter of the stimulus aperture was ~60 µm. (b) and (c) Intrinsic optical signals elicited by a 125 ms visible light flash. Each illustrated frame is an average over 100 ms interval (8 frames). 100 ms pre-stimulus and 500 ms after-stimulus images are shown in each imaging sequence. The imaging sequence c was recorded after the imaging sequence b, with a time interval of 2 minutes. (d) Temporal change of intrinsic optical responses. The numbered tracings 1–6 are corresponding to the arrows pointed retinal areas (average of 5×5 pixels) in sequence b. Vertical lines indicate the stimulus onset and offset. (e) (Media 1) video showing dynamic intrinsic optical signal patterns at the photoreceptor layer. This video is from the same image sequence shown in b. 0.5 s (40 frames) pre-stimulus baseline and 5 s (400 frames) after-stimulus images are shown.

Download Full Size | PDF

At the photoreceptor layer, the distribution of major IOSs showed a negative-positive (center-surround) hollow structure. The retinal area covered by the stimulus light spot was dominated by negative signals, although positive-going changes were also observed, particularly at early phase of the retinal response, such as arrows 1 and 2 pointed areas in Fig. 2(b). Robust positive IOSs were observed in the surrounding area of the stimulus spot. Fast IOSs were initiated in the center of the stimulus spot first, and spread to the surrounding area rapidly (Fig. 2(e)). In the center area, the rapid positive-going IOSs (tracings 1 and 2 in Fig. 2 (d)) initiated immediately after the stimulus onset, and reached the amplitude peak within 50–100 ms. Most of the negative and positive responses reached an amplitude peak within 200 ms after the stimulus onset, but the recovery phase lasted a relatively long time, typically a few seconds (Fig. 2(e)). From Fig. 2(d), we can see that the amplitude peaks (or flexural points) of fast IOSs have time courses comparable to ERG a- and b-waves.

 

Fig. 3. (a) Representative CCD image sequence of inner nuclear layer, without differential processing. The white spot in the second frame shows the visible stimulus pattern. At the photoreceptor layer, the diameter of the stimulus aperture was ~60 µm. (b) and (c) Intrinsic optical signals elicited by a 125 ms visible light flash. Each illustrated frame is an average over 100 ms interval (8 frames). 100 ms pre-stimulus and 500 ms after-stimulus images are shown in each imaging sequence. The imaging sequence c was recorded after the imaging sequence b, with a time interval of 2 minutes. (d) Temporal change of intrinsic optical responses. The numbered tracings 1–6 are corresponding to the arrows pointed retinal areas (average of 5×5 pixels) in sequence b. Vertical lines indicate the stimulus onset and offset. (e) (Media 2) video showing dynamic intrinsic optical patterns at the inner nuclear layer. This video is from the same image sequence shown in b. 0.5 s (40 frames) pre-stimulus baseline and 5 s (400 frames) after-stimulus images are shown.

Download Full Size | PDF

The center-surround distribution of positive and negative IOSs at inner retinal layers was opposite to that of the photoreceptor layer. Strong optical responses with positive polarity were observed in the stimulus area, and negative signals were mainly distributed at the surrounding area. The optical patterns of the IOSs were irregular (compared to the circular stimulus aperture) (Figs. 3–4). The irregular patterns of the IORs may reflect localized activities of retinal neurons, with nonsymmetrical distribution relative to the stimulus center. From Figs. 3–4, we observe that the amplitude peaks of the IOSs at the inner retina also have time courses comparable to ERG a- and b-waves.

 

Fig. 4. (a) Representative CCD image sequence of ganglion layer, without differential processing. The white spot in the second frame shows the visible stimulus pattern. At the photoreceptor layer, the diameter of the stimulus aperture was ~60 µm. (b) and (c) Intrinsic optical signals elicited by a 125 ms visible light flash. Each illustrated frame is an average over 100 ms interval (8 frames). 100 ms pre-stimulus and 500 ms after-stimulus images are shown in each imaging sequence. The imaging sequence c was recorded after the imaging sequence b, with a time interval of 2 minutes. (d) Temporal change of intrinsic optical responses. The numbered tracings 1–6 are corresponding to the arrows pointed retinal areas (average of 5×5 pixels) in sequence b. Vertical lines indicate the stimulus onset and offset. (e) (Media 3) video showing dynamic intrinsic optical patterns at the ganglion layer. This video is from the same image sequence shown in b. 0.5 s (40 frames) pre-stimulus baseline and 5 s (400 frames) afterstimulus images are shown.

Download Full Size | PDF

With consistent stimulation protocol and a fixed recording area, sequential measurements disclosed IOSs with a reproducible pattern in a selected retinal layer (Figs. 2–4). For different retinal samples (or different recording areas), the signal amplitude and pattern of the IOSs were variable, although the stimulus was consistent. However, we typically observed consistent center-surround distribution of the positive and negative IOSs. Fig. 5 shows averaged images of IOSs recorded from 6 isolated frog retinas. In order to ensure signal reproducibility, 2 imaging passes were implemented at each (i.e., photoreceptor, inner nuclear, and ganglion) layer of the 6 retinas, with consistent recording area and stimulus strength (duration: 125 ms; intensity: 4.0×10⁴ [550-nm photons]/ms·µm^-2). In other words, each image sequence shown in Fig. 5 is an average of 12 experimental trails.

The center-surround distribution patterns of major IOSs were consistently observed from the averaged intrinsic optical signal images of the photoreceptor and inner retinal layers. The photoreceptor layer was dominated by negative optical signals, although positive signals were also observed, particularly at the surrounding area (Fig. 5a). At the inner nuclear and ganglion layers, negative responses were mainly observed at the surrounding area; while strong positive signals were mainly confined in the center area. In comparison with the inner nuclear layer, the negative responses in ganglion layer spread to a relatively larger area, but with lower signal amplitude.

 

Fig. 5. (a–c) Averaged intrinsic optical images of the photoreceptor (a), inner nuclear (b) and ganglion (c) layers. The white spot in the first frame of sequence (a) shows the stimulus pattern. At the photoreceptor layer, the diameter of the stimulus spot was ~60 µm. Each image sequence is an average of 12 experimental passes, and each illustrated frame is an average over 100 ms interval (8 frames). 100 ms pre-stimulus and 500 ms after-stimulus images are shown in each imaging sequence. (d–f) Statistics of positive and negative optical responses at the photoreceptor (d), inner nuclear (e), and ganglion (f) layers, corresponding to the areas marked by the square blocks shown in the third frames of sequences a–c.

Download Full Size | PDF

Fig. 5 (d–f) shows the statistics of retinal area with positive (>0.3% ΔI/I) and negative (<-0.3% ΔI/I) responses. A threshold (0.3% ΔI/I) was used to reduce the effect of background noises on the statistics. From Fig. 5 (d–f), we can see that the stimulus-activated retina area with negative optical changes consistently approach an amplitude peak within 200–300 ms after the stimulus onset, although the amplitude peak degrade from photoreceptor to ganglion layer. For the stimulus-activated retinal area with positive optical signals, the amplitude peaks are at 500–1000 ms after the stimulus onset. The ganglion layer shows the largest amount of retinal area with positive optical signals.

3.2. CMOS imaging of ON and OFF responses

Using the fast CMOS camera, we characterized IOSs associated with ON and OFF edges of a visible light stimulus. Figure 6 represents IOSs elicited by a 500 ms step stimulus. The stimulus light intensity was 1.0×10⁵ [550-nm photons]/ms·µm^-2. The raw CMOS images were captured with an imaging speed of 1000 frames/s, and thus provided a 1 ms temporal resolution to characterize the ON and OFF responses.

 

Fig. 6. (a–c) Representative CMOS images of the photoreceptor (a), inner nuclear (b), and ganglion (c) layers of the retina before differential processing. Each CMOS frame consists of 160×120 pixels. The left-top and bottom right corners are Pixel (0, 0) and Pixel (159, 119), respectively. The white spot in (a) shows the stimulus pattern. At the photoreceptor layer, the diameter of the stimulus spot was ~60 µm. (d–f) (Media 4) Video clips of dynamic intrinsic optical changes recorded from the photoreceptor (d), middle (e) (Media 5), and ganglion (f) (Media 6) layers. The retina was activated by a 500 ms step light stimulus. The raw CMOS images were recorded with a speed of 1000 frame/s. Each illustrated frame of the videos is an average over 10 ms interval (10 frames). 200 ms pre-stimulus and 1000 ms after-stimulus images are shown in each video. The imaging sequences d, e, and f were recorded sequentially, with a time interval of 5 minutes. (g) Transient intrinsic optical changes of individual pixels. In each subpanel, Pixel (x, y) indicates the location of each representative pixel. Black, blue, and red tracings correspond to the intrinsic optical signal images recorded from photoreceptor layer (PRL), inner nuclear layer (INL), and ganglion layer (GL). Vertical lines indicate the stimulus onset and offset. Black arrows point to ON response of the IOSs corresponding to the stimulus onset. Green and purple arrows point to OFF response of the IOSs corresponding to the stimulus offset.

Download Full Size | PDF

Consistently, we observed both positive and negative IOSs from each retinal layer. Strengthened stimulus disclosed enhanced center-surround contrast of the IOSs with opposite polarities, particularly at inner retinal layers. Fast NIR light images disclosed both ON and OFF responses of the IOSs associated with stimulus onset and offset, respectively. Figure 6 (g) represents transient intrinsic optical responses of individual pixels. Fast IOSs typically reached the amplitude peak (black arrows) within 50–200 ms after the stimulus onset. A flexural point (green arrows) was observed at the OFF edge of the stimulus, and an additional amplitude peak (purple arrows) was also occasionally observed after the stimulus offset. Figure 7 shows averaged IOSs of the center retinal area marked by a white square in Fig. 6 (b). With improved SNR of the averaged optical signals, we observed that the on-going phase of the IOSs began within 5 ms after the stimulus onset (Fig. 7), and both ON (black arrows) and OFF (green arrows) responses were consistently observed, particularly at the inner retinal layers.

 

Fig. 7. Averaged intrinsic optical responses of the retinal area (40 µm×40 µm) covered by the white block shown in Fig. 6 (b). Black, blue, and red tracings correspond to the imaging sequences of the photoreceptor layer (PRL), inner nuclear layer (INL), and ganglion layer (GL) shown in Fig. 6 (d–f). Vertical lines indicate the stimulus onset and offset. Black arrows point to ON response of the IOSs corresponding to the stimulus onset. Green and purple arrows point to OFF response of the IOSs corresponding to the stimulus offset.

Download Full Size | PDF

4. Discussion

In summary, a flood-illumination NIR light microscope was used to conduct depth-resolved enface imaging of IOSs in isolated frog retina. Fast IOSs initiated within 5 ms after the stimulus onset, and typically reached an amplitude peak within 50–200 ms. Prolonged (500 ms) step stimulus disclosed ON and OFF responses corresponding to the stimulus onset and offset, respectively. The time courses of fast IOS were comparable to ERG response. Both positive and negative IOSs were observed at the photoreceptor and inner layers. Depth-resolved images disclosed different center-surround distributions of the positive and negative IOSs at the photoreceptor and inner retinal layers. At the photoreceptor layer, negative IOSs dominated the center area illuminated by the stimulus light spot, while positive IOSs dominated the surrounding area. In contrast, at inner nuclear and ganglion layers, positive IOSs dominated the center area covered by the stimulus light spot, and negative IOSs were mainly observed in the surrounding area.

It is well known that the ERG is an integral measurement of electrophysiological responses over the whole depth of the retina. Similarly, the IOSs may also consist of contributions from multiple retinal layers and cells. Different center-surround distributions of the positive and negative IOSs at the photoreceptor and inner retinal layers support the hypothesis that the negative IOSs are mainly associated with photoreceptor response and the positive IOSs result primarily from dynamic changes of post-photoreceptor response during retinal activation [16]. In principle, when the NIR probe light is focused at the photoreceptor layer, intrinsic optical signal image should be dominated by the IOSs produced by the photoreceptors. Although cross-talk response (defocused transmitted or scattered light) of the IOSs from other layers may also enter the CCD/CMOS imager, these light photons should be spread (blurred) into a relatively larger area. In comparison, the IOSs directly initiated from the photoreceptor layer are in focus. On the other hand, when the NIR probe light is focused at the inner retinal layers, the intrinsic optical signal image should be dominated by the IOSs initiated at the corresponding retinal depth. Thus, the positive IOSs in the surrounding area of the photoreceptor layer may result from the defocused IOSs associated with the responses of inner retinal neurons; while the negative IOSs in the surrounding area of the inner retinal layers may result from the defocused IOSs associated with photoreceptor response. In comparison with the inner nuclear layer, the negative IOSs at the ganglion layer spread into a relatively larger area, but with decreased signal amplitude. This agrees with the fact that the ganglion layer has a relatively larger distance from the photoreceptor layer, and thus the light from the photoreceptor layer may be defocused (blurred) over a relatively larger area. Moreover, the optical waveguiding property of photoreceptors [27, 28] and other cells [29] may also produce cross-talk of the IOSs among different retinal layers. The optical waveguiding property of the retina has directional dependence, i.e., optical Stile-Crawford effect [30], and thus may also affect the center-surround distribution of the IOSs.

From the above discussion, we hypothesize that the major part of the positive IOSs observed at the photoreceptor layer, particularly at the retinal area outside of the stimulus light spot, may result from cross-talk IOSs from inner retina. However, both positive and negative IOSs, particularly inside the stimulus area, may also relate to phototransduction procedures. Previous studies with isolated photoreceptor outer segments and isolated retinas have demonstrated transient IOSs associated with phototransduction [3, 31, 32]. Both binding and release of G-proteins to photoexcited rhodopsin may contribute to the IOSs [31]. While the negative IOSs may result from the binding of photoexcited rhodopsin to G-proteins; some part of the positive IOSs may relate to the dissociation of the complex upon GDP/GTP exchange [31]. The transient optical patterns of the IOSs at the photoreceptor layer may partially result from antagonistic contributions of the negative and positive IOSs associated with dynamic redistribution of G proteins during retinal activation. Further study is required for investigating possible effect of center-surround antagonism [33] of the retina on the observed center-surround distribution of IOSs with positive and negative polarities.

Although biophysical mechanisms of the observed fast IOSs at inner retinal layers are not well known, previous studies with eyecup slices have demonstrated transient scattering changes at inner retina [6]. The studies with other neural tissues have suggested several possible processes [34], such as neurotransmitter secretion [35], reorientation of membrane proteins and phospholipids [7, 8, 36], and refractive index change of neural tissues [37], to produce transient IOSs during neural activation. Our recent experiments and theoretical analysis support that the fast IOSs may, at least partially, result from dynamic volume changes of activated neurons corresponding to ion and water flow across the cell membrane [38, 39]. Water influx in response to ionic currents through gated channels during depolarization causes cellular swelling that can produce changes in tissue light scattering [38–41], and also polarization changes [42].

Further investigations are required for deeper understanding and better recording of fast IOSs associated with retinal activation. While transmitted light imaging of isolated retinas can provide important application for study of visual information processing in the retina, reflected light imaging is typically required for pursuing in vivo detection of fast IOSs. By constructing a home-built confocal microscope, which has sub-millisecond temporal resolution and micrometer spatial resolution, we have recently demonstrated the practicality of reflected light imaging of fast IOSs associated with retinal activation. Consistently, we observed both negative and positive optical signals, with time courses comparable to ERG kinetics. In fact, we [15] and other groups [13, 23] have previously demonstrated depth-resolved reflected light recording of stimulus-evoked IOSs in isolated retina [13, 15] and living eye [23] using fast functional OCTs. However, in vivo depth-resolved OCT recording of fast IOSs, which have time courses comparable to ERG responses, has been challenged by the limited SNR, probably due to speckle noise of coherent illumination light in the OCT system. Further improvements of advanced optical imaging techniques, such as a high-speed OCT imager with improved SNR or an adaptive optics imager with improved depth resolution, will allow reliable measurement of fast IOSs. We anticipate that high spatiotemporal resolution imaging of fast IOSs, which have time courses comparable to ERG responses, will advance the study of dynamic visual processing in the retina, and also provide new methodology for early detection of retinal diseases and reliable assessment of treatment outcomes.