Selective Retina Therapy
Laser photocoagulation of the retina has been performed for more than 40 years and has enabled the denaturation of retinal layers by the deposition of heat in tissue. However, for treatment of pathologies associated retinal pigment epithelium (RPE), excessive damage of adjacent tissue, especially the damage of the photoreceptor layer is considered to be disproportionate.
Selective Retina Therapy (SRT) using microsecond laser pulses specifically targets RPE cells while sparing the surrounding tissue, to prevent long term damage, scarring and scotoma. Different approaches for control of minimally invasive treatment have been pursued. However, since SRT effects are ophthalmoscopically invisible a device for high precision application, documentation as well as a dosing control is required.
Coaxial integration of a 532 nm μs-pulsed treatment laser enhanced the clinical diagnostic SPECTRALIS system (Heidelberg Engineering) to become a multimodal treatment platform for image-guided coagulation and selective retina therapy.
This combined optical coherence tomography (OCT)-SRT system provides precise pre- and post-therapeutic diagnostics for planning and supervision. Furthermore, it has the potential to limit the laser energy deposition used locally for individual treatment via detection of microbubble formation to protect the neuronal retina from overexposure and collateral damage during selective removal of the retinal pigment epithelium.
KeywordsSelective retina therapy Selectivity Subthreshold treatment Retinal pigment epithelium (RPE)
11.1 Retinal Therapy: A Short Historic Overview
Retinal light based therapy has been discovered in the late 1940s by using sunlight  which was soon replaced by Xenon light flashes, and eventually by coherent monochromatic radiation, once powerful lasers became available. The latter first proved successful coagulation in clinical trials in the early 1970s .
Traditional laser photocoagulation (LPC) utilizes thermal interaction by which light energy absorbed by tissue pigments is converted to heat, thus causing photothermal denaturation whereas the temperature rise is dependent on laser power, wavelength and exposure time, in addition to the properties of the target tissue . LPC techniques have been a mainstay for the treatment of several retinal diseases. Multiple hypotheses regarding the biological effects of the different diseases are still in discussion , including reduction of oxygen demand by destroying photoreceptors, alterations of the local permeability and laser stimulation of RPE in edematous areals for proliferation and rejuvenation. Independent of the lack of a full understanding of the underlying mode of action, LPC was found beneficial to prevent retinal detachment after retinal hole formation , for the treatment of diabetic retinopathy (DR) , diabetic macular edema (DME) , neovascular age-related macular degeneration (AMD) , for central serous retinopathy (CSR)  and some other retinal pathologies.
LPC has been shown to be very successful in panretinal photocoagulation for DR hypothesizing that a reduction in oxygen demand by coagulating retinal tissue in the periphery preserves the central vision. Further, a macular grid treatment for DME was and still is very successful with respect to maintaining vision and reducing oedema in the central macula.
Laser therapy for AMD has an even more changeful history that was driven by the availability of laser technology and diagnostic tools. Starting in the 1970s, laser parameters used in early studies up to the late 1990s and even at the beginning of the new millennium were widely inconsistent . Reports on improvements as well as lack of correlation in respect to the treatment can be found. In most cases, researchers utilized readily available laser diodes or argon ion lasers with the ability to irradiate tissue in continuous or quasi continuous wave (cw) irradiation modes. As most recently reviewed in  studies involved fellow-eye treatment of drusen sites and their short-term visibility as well as long-term changes. Indicating either slightly more positive or negative outcomes for AMD, the inconsistency in the results might be associated with different levels of choroidal neovascularization due to inflammatory response caused by necrosis of RPE and neighbouring tissue. LPC posed an important tool before the advent of a wide range of pharmaco-therapeutical agents that diminished the significance of laser-treatment to time-critical cases. Despite the drugs’ success, several clinical trials indicate that therapeutic effects of intravitreal agents such as anti-VEGF injections and steroid implants are short-term compared to those of traditional laser photocoagulation therapy (LPC) and furthermore have the drawback that their application cannot be precisely controlled spatially . To avoid regular arduous, cost-intensive and also risky injections required for chemical treatment of the chronic and recurrent disease, improvements in laser irradiation were pursued.
Due to the slow heating process and the strong heat dissipation into the surrounding tissue when using millisecond laser exposure, massive collateral damage of surrounding components like Bruch’s membrane and choriocapillaris as well as the particularly healthy overlying photoreceptor (PR) cells is unavoidable [10, 11]. These adverse effects can lead to scotoma, reduced night vision, and disruption of the retinal anatomy through scarring [12, 13].
In the last decade, microsecond pulsed lasers in the green spectral range have become readily available for clinical applications that try to induce a specific therapeutic effect, but avoid damage of the PR, neural retina and choroid by selective targeting of different fundus structures. The following sections describe the laser effects on the RPE layer, the idea of its selective treatment, the underlying physics as well as initial experiments and clinical study results.
11.2 The Concept and State of the Art of Selective Retina Therapy
For a variety of retinal diseases, especially those which are thought to be associated with a degradation and reduced metabolism at the retina, it might be elegant to selectively eliminate specific regions of the monocellular RPE layer and thereby trigger its regeneration by wound healing. It is known that the RPE is able to rapidly close small wounds as found after mild photocoagulation. Such a cellular rejuvenation might lead to an improved metabolism, molecular transport and retinal functionality. However, the adjacent highly sensitive neural retina and choroid should not be affected by such a process.
The question arose if and how such a selective RPE treatment can be induced. In 1983 Anderson and Parrish published the idea of selective photothermolysis based on the concept of an absorption contrast with a stronger light absorption in the target area embedded in a less absorbing surroundings . Selective treatment is postulated when using repetitive laser pulses with peak temperatures allowing sufficient denaturation rates just at the absorbing sites. This concept is perfectly suited for the RPE cells being heavily loaded with strongly light absorbing melanosomes (Fig. 11.1: 100–200 per cell, each about 1 μm in size). About 20–80% of light in the green spectral region is absorbed by the RPE . Melanosome distribution and absorption varies inter- and intraindividually. It typically increases towards the periphery and decreases with age .
In order to selectively heat and damage the RPE only, without affecting the surroundings, heating should take place far below the RPE’s thermal confinement time, defined as decreasing of the maximum temperature rise to 36% (1/e). When assuming a 3 μm uniformly absorbing sheet, neglecting the individual melanosomes, and a much larger heating spot diameter, then about 30 μs are estimated. Thus using heating times far below 30 μs should enable strong temperatures at the RPE with strong temperature gradients towards the much less absorbing environment . Such a quick heating can be obtained with high power laser pulses, preferentially at wavelengths in the green spectral range due to high melanin absorbance. Unfortunately, μs heating time requires very high irradiance and thereby laser power in order to obtain temperatures for a sufficient thermal damage rate. According to the Arrhenius theory describing thermal damage as a first order rate process , a 30 μs single pulse already requires temperatures exceeding 100 °C, when taking the commonly used Arrhenius constants for retinal denaturation into account. However, as thermal damage is accumulative, applying multiple pulses with lower damage rates was assumed to be also sufficient, provided that an appropriate cooling time between the pulses is met in order not to lose selectivity owing to background heating by thermal diffusion.
11.2.1 Experimental Results
As an additional concept, a continuous wave argon ion laser beam was used in a rapid scanning mode also providing μs illumination times. Using an 19 μm laser spot diameter at a scanning speed of 11.7 m/s in order to achieve a 1.6 μs irradiation time, a threshold power of 569 mW was found for ten repetitive scans, which corresponds to a threshold radiant exposure of 297 mJ/cm2 . Thresholds found here were slightly higher than with the pulsed application, but also showed saturation for multiple hundred scans (Fig. 11.2) . The technique was approved on rabbits showing the desired RPE effects . Interestingly, the retinal spot scanning speed of Heidelberg Engineering’s confocal scanning laser ophthalmoscopes (cSLO) is around 70 m/s which corresponds to a single spot irradiation time of about 270 ns. This irradiation time had already been proved with a pulsed laser in clinical trials  and thus a cSLO represents another promising irradiation modality for SRT.
However, the threshold dependence on the pulse duration and the saturation with higher number of pulses found were in contradiction to the thermal damage theory of Arrhenius. Thus other mechanisms of cell damage had been taken into account. Lin et al. showed selective damage of artificially pigmented cells by microbubble formation (MBF) around phagocytosed absorbers induced by Q-switched laser pulses . Microbubble occurrence always coincided with immediate cell death observed with a vitality stain. In order to understand MBF closer, the melanosomes and the granular structure of the absorbers needed to be taken into account. The thermal confinement time for a 1 μm diameter melanosome can be estimated to about 450 ns . Microbubbles nucleate heterogeneously at the surface of the melanosomes and then grow around the melanosomes . Investigations on the bubble dynamics revealed a growth of the bubbles proportional to the radiant exposure for ns pulse duration, but an oscillation with limited maximal size of about 4 μm in a self-limiting process when using μs pulse durations . Investigating the size of the microbubbles around single melanosomes by using fast flash photography scales proportionally with their lifetime according to the Rayleigh equation , which is also valid for cavitation bubbles. Nucleation temperatures have been experimentally found to be around 136 °C for ns  and 157 °C for 1.8 μs pulse duration , respectively. These temperatures correspond to the nucleation temperature of water under a pressure of 3–4 bar, which also re-presents the surface tension of a microbubble that needs to be overcome for growing. Temperature and Arrhenius calculations using a mash of melanosomes interestingly also showed that the threshold dependence experimentally found with respect to the pulse duration is only applicable at the surface of the melanosomes .
Experiments on rabbits in vivo revealed that RPE disruption is ophthalmoscopically invisible with white light illumination during treatment . However, it can be demarcated by fluorescence angiography. In case of tight junction disruption within the RPE, fluorescein can diffuse through the broken blood retina barrier and fluorescing spots are observed at these locations. Typical ED50 threshold radiant exposures for μs pulses (100 pulses with a single pulse duration of 1.7 μs) are determined by angiography to 131 mJ/cm2, and therefore are comparable to those ones found on porcine RPE explants ex vivo labelled by a vitality stain. At radiant exposure about twice the ED50 RPE damage threshold, effects become optical visible . They appear like mild whitish photocoagulations, but can most likely be attributed to increased light scattering due to the mechanical dislocation and disruption of retinal tissue by large coalesced macrobubbles. With further increase of the radiant exposure, these macrobubbles lead to retinal/choroidal disruption with bleedings as typically observed after Q-switched laser accidents. For a safe and selective treatment, a therapeutic window with radiant exposures slightly above the MBF threshold and far below twice the MBF threshold should be considered. When investigating the cellular damage mechanisms on pulse duration close to threshold irradiation in more detail, it shows that the predominant origin is thermomechanical for pulse durations up to 20 μs and thermal for pulse durations of 50 μs and larger [36, 37].
Cytokine release from RPE also alters after irradiation; an investigation with ex vivo RPE showed a decrease of the VEGF and an increase of the PEDF secretion from RPE cells 3 days after irradiation . Proinflammatory cytokines, including IL-1β and TNF-α, showed significantly lower levels after SRT than in LPC . It has also been shown that matrix metalloproteinases (MMPs), particularly the active form of MMP-2 is significantly upregulated [39, 40, 44] and also locally resolved as shown in Fig. 11.5 , which is assumed to play a positive role in preventing matrix degradation of Bruch’s membrane, and eventually increasing transmembrane molecular exchange. Human donor samples of Bruch’s membranes incubated with MMP-2 showed considerably improved hydraulic conductivity and transport capabilities [17, 40]. Interestingly, the upregulation of active MMP-2 and PEDF in the whole RPE explant are more apparent with 200 μm spots compared to 100 μm spots when the total damage area was kept constant . This suggests that the spot size of the RPE damage may also have an influence on the subsequent cytokine release and thus may affect the therapeutic effect. In conclusion, RPE cell regeneration together with an improved molecular diffusion rate through Bruch’s membrane might contribute to the therapeutic effectiveness, as the complete choroidal-retinal complex is affected by SRT.
11.2.2 Clinical Study Results
Lowest angiographically visible radiant exposures were found in the range of 350–500 mJ/cm2 per pulse in test expositions at the arcades, which is about 2.7–3.8 times higher than the rabbits angiographic ED50. For treatment, typically 650 mJ/cm2 were used to compensate for the variation in pigmentation, which is typically lower in the central macula. It quickly turned out that the repetition rate of 500 Hz is too high for a spot diameter of 160 μm in order to allow sufficient cooling between the pulses. Successively, until today 100 Hz is the standard repetition rate used.
The reason for the much higher required radiant exposures compared to the very young rabbit and pig is very likely related to the higher light scattering within the patient’s eye before the radiation can reach the retina. Boettner and Wolter measured a strongly reduced direct light transmission through human eyes with age . Only about 40% direct light transmittance in the green spectral range was measured for a 53 year old eye. This value further reduces with age owing to increased light scattering predominately in the lens due to age-related cataract, therefore well explaining the higher thresholds for clinical SRT. These findings also coincide with temperature measurements during retinal photocoagulation: Only about 20–30% of the green light entering the eye reached the target site at the retina and led to the temperature rise measured by optoacoustics in patients .
With respect to the therapeutic outcome in the follow-up period, hard exudates disappeared in 6 out of 9 patients, retinal edema resolved in 6 out of 12 diabetic patients, drusen were reduced in 7 out of 10 AMD patients and leakage in CSR disappeared in 3 out of 4 cases . Investigations of SRT compared to photocoagulation lesions directly after treatment utilizing optical coherence tomography (OCT) revealed no OCT-visible hyperreflectivity at the SRT sites in contrast to the strongly scattering photocoagulation spots . In both cases, the RPE appeared thinner in the follow-up. After 4 weeks, RPE thickening likely due to overproliferation was observed after SRT. One year after treatment, photocoagulation sites were characterized by RPE and neurosensory tissue atrophy, in contrast SRT lesions showed unaffected neurosensory structures and an intact RPE layer.
Encouraged by these promising pilot study results, an international SRT multicenter trial was conducted to evaluate the therapeutic effect of SRT in 60 patients with diabetic maculopathy . In this study, 30 pulses with a single pulse duration of 1.7 μs at 100 Hz were applied per spot with a top hat beam profile of 210 μm on the retina. A Nd:YLF laser with intracavity overcoupled second harmonic generation was developed to generate μs laser pulses at 527 nm . Typical pulse energies for DME treatments were 200–325 μJ, which refer to pulse powers between 117 and 191 W and calculated radiant exposures between 577 and 938 mJ/cm2, respectively . In 95% of the 60 patients the visual acuity improved or was stable at the 6 month follow-up control. The angiographic results with respect to leakage areas and OCT on edema size and thickness, however, didn’t fully coincide with respect to the visual acuity. Further studies on DME , with macular edema after venous branch or central venous occlusion  and with geographic atrophy , were conducted.
Most impressive are study results for the treatment of central serous retinopathy (CSR). A study on 27 patients with active CSR showed a resolved subretinal fluid in 85.2% of the patients after 4 weeks and 100% after 3 months with no visible leakage. Mean visual acuity improved from 20/40 at baseline to 20/20 at 3 months . A study randomizing patients to immediate treatment and 3 month waiting time also showed significant benefits and quick fluid dissolution for the patients treated immediately . A study conducted in Japan supported these findings . With respect to the safety of the treatment, microperimetry and multifocal ERG did not show any functional retinal defects, as typically found after standard photocoagulation . Another pilot study conducted in Korea (527 nm, 100 Hz, 30 pulses per spot, 210 μm diameter on retina) on chronic (>3 months) CSR patients revealed much lower pulse energies between 65 and 90 μJ needed for selective RPE effects . Angiographically visible effects were observed with as little as 70 μJ (200 mJ/cm2). Reasons for these lower radiant exposures are likely the relatively clear media of the younger CSR patients, the higher pigmentation of Asian eyes as well as the very strong intensity modulation on the lateral laser beam profile with that laser used. However, only 3.8 laser spots in average per patient were applied around the leakage point leading to a complete resolution of subretinal fluid in 75% of the patients after 9 months. Only very few spots around the feeder vessel need to be applied in the central macula, in contrast to a study using a non-damaging laser therapy using 548 laser spots in average per patient . Thus for CSR, SRT seems to be an ideal treatment modality.
Another clinical study investigated shorter pulse durations between 200 and 300 ns for SRT, which also exhibited selective RPE damage and even lower pulse energies compared to 1.7 μs due to the reduced heat diffusion during irradiation [25, 61]. However, since the individual nucleation times of the multiple microbubbles converge, the dynamics becomes more explosive and the risk of unintended bleedings rises, as clinically observed following the impact of single 3 ns laser pulses when using another technique for retinal rejuvenation .
The primary disadvantage during SRT treatment is the lack of direct optical feedback due to the low or missing visibility of SRT lesions. On the one hand, it strongly limits the ophthalmologist ability to interactively place the lesions in the desired areas. Neither reference grids for DME nor well localized circles or patterns in CSR treatment can be placed due to lack of visual feedback. On the other hand, with respect to proper energy dosage, studies showed a strong inter- and intraindividual variation of the necessary pulse energies for comparable RPE effects [63, 64]. Even with titration at the arcades, angiography after treatment is required to ensure RPE defects due to the variability of the absorption. In case of negative angiography outcome, retreatment with higher energies is required, which is a tedious procedure with increased risk for the patient owing to multiple angiographies. The state of the art to improve and control dosage will be discussed in the next paragraph.
11.2.3 Dosimetry and Dosing Control
Small bandwidth interferometry
Broad bandwidth interferometry-OCT
With respect to clinical applications within the approval phase of such technique, currently the physician is guided by the system as follows: The clinician chooses the maximal energy of the last pulse and the system applies a stepwise ramp beginning with 50% of the maximal pulse energy. MBF is measured after each single pulse. After all 15 pulses have been applied, the system proposes to keep the energy for the next spot if MBF is measured with pulses in the middle of the ramp. In case of MBF at the beginning or end of the ramp, the physician is asked to decrease or increase the maximal energy, respectively [54, 71].
Light absorbed by tissue is predominately converted to heat, which leads to its thermo-elastic expansion and successive contraction by the emission of a bipolar pressure wave . Its frequency depends on the thickness of the absorber and the spot size, while its amplitude is proportional to the pulse energy and the strongly temperature dependent Grüneisen parameter. Therefore optoacoustics can be used to calculate the tissue’s temperature increase during SRT  and also during standard photocoagulation . Typically, frequencies after heating tissue with ns to μs pulses are in the MHz and therefore in the ultrasonic range. The acoustic waves travel back through the eye and can be measured at the cornea. As the pressure amplitudes are in the range of hundred microbars only, a very sensitive sensor is required. Typically, an annular shaped piezoelectric ultrasonic transducer is embedded in the contact lens, which is used anyway for laser treatments utilizing a laser slit lamp . Figure 11.7b (blue curves) show such transients when using SRT at low radiant exposures . Since the pressure is proportional to the pulse energy, all transients are almost identical when normalizing on the energy (Fig. 11.7b). If the temperatures exceed the vaporization temperature at the surface of melanosomes, microbubbles form and expand [74, 75], and successively emit further individual pressure waves. As the nucleation onset times at the different melanosomes differ slightly, and also differ among different pulses owing to the displacement of the melanosomes by the bubbles, a phase and amplitude fluctuation of the transients is observed as exemplary shown in Fig. 11.7b (green and red curves). With an algorithm developed for taking into temporal deviations among the acoustic transients, Schüle et al. introduced an optoacoustic value (OA value) . With an appropriate threshold OA-value, RPE damage upon fluorescein angiography can be predicted with high specificity [76, 77]. Also other algorithms based on further signal features can be processed, giving a higher security for the feedback when processed in parallel . An appropriate threshold OA-value can be used in real time to interrupt the laser ramping and emission after MBF exceedance. The beauty of this approach is that the excitation is induced by light while the response is based on acoustics, since the pressure waves are almost not affected by the ocular media. From the perspective of bubble formation, the correct focus of the light or the beam profile itself play only very minor roles. The technique has been proven in clinical studies by displaying the OA-value to the ophthalmologist via a monitor [52, 63] or verbally  in order to give them immediate feedback whether the irradiation energy was sufficient. In recent clinical studies, it was also used for dosing guidance [54, 71].
During laser irradiation of the retina, some of the treatment light is scattered back from the RPE and can be measured in front of the eye. If only tissue heating takes place, its reflective properties stay almost constant and the pulse shape of the back-reflected light is identical to the incoming one (Fig. 11.7c). After MBF however, microbubbles cause a refractive index change at the bubble’s surface between tissue on the outside, essentially consisting of liquid water (n = 1.4) and water vapor inside the bubble. (n = 1) of about 2–3 orders of magnitude. Further the light is modulated owing to the dynamics of the bubble expansion and collapse. Figure 11.7c (green and orange curves) shows such back-reflected pulse shapes which strongly deviate from those below bubble-formation threshold. This method was developed by Seifert et al.  with respect to calculate a reflectometry value (RE-value). Data from clinical trials were recently analyzed and mathematically optimized in order to obtain a threshold RE-value with highest sensitivity and specificity  for an automatic real-time stop of the pulse energy ramp. The technique is currently under investigation in clinical trials [54, 64, 71, 77]. The strength of this method is that the high power laser light used for treatment can be evaluated directly. Therefore on the hardware side, nothing but a fast photodiode is required to record the strong backscattered light. However, in contrast to optoacoustics, the reflected light intensity is influenced by the ocular media as lens and cornea and can strongly vary as also the backscatter of a range of small spherical ‘bubble-foam’ is highly dependent on the particular microbubble distribution.
Small spectral bandwidth interferometry
The occurrence and dynamics of microbubbles can be well probed by interferometry using small bandwidth laser light with large coherence length as typically provided by single longitudinal mode lasers. By means of a fast photodiode collecting the reflected light from the eye in the object arm of the interferometer, microbubble dynamics can be recorded at very high temporal resolution and thus provide more information than just the MBF threshold detected by OA- and RE-values. With a single frequency laser diode, the heating of the tissue in form of a slight intensity shift due to the thermal expansion of the RPE, as well as the onset of microbubbles (modulations in Fig. 11.3), their coalescence and collapse, and finally the slow thermal contraction after irradiation have been monitored [31, 32]. Thus small bandwidth interferometry can likely be used for fine dosing control in SRT , however has not been developed far enough for approved clinical trials.
Broad spectral bandwidth—OCT
Low coherence or white light interferometry (LCI) responds similarly to reflectivity changes like any of the other optical approaches. The broad bandwidth and the coherence between the different wavelengths, however, can also be utilized to transmit spatial information of the reflector in the light propagation direction. Optical coherence tomography combines this technology with a lateral scan and enables to obtain images of reflectivity in form of axio-lateral cross-sections or volumes. The availability of performing OCT during retinal therapy exposure for observing signal variations correlated to the creation of RPE lesions has already been investigated at times where early frequency-domain OCT system have become widely available . However, at that time the temporal resolution was only sufficient to monitor thermal expansion of tissue. For real-time diagnostics of laser-tissue interaction, two time-frames that require quite different acquisition speeds have to be distinguished. One is the thermal expansion regime that is specified by thermal diffusivity in the range of up to 0.16 mm2/s for biological tissue , leading to speeds of ~80 mm/s and the other is where acoustic waves with ~1500 m/s dominate the energy transfer to surrounding tissue . Several groups have investigated structural OCT as an imaging modality to compare retinal layers before and after treatment [51, 82], or functional Doppler OCT for thermal expansion measurements [79, 83]. This direct imaging of thermal expansion during the sufficiently slow heat diffusion process is straight forward with current OCT systems at moderate axial resolutions ~10 μm and ~100 kHz depth scan rates even for the lateral cross-sections of OCT tomograms. Although modern OCT systems can scan remarkably faster (>1 MHz depth scan rate, being <1 μs exposure time per depth scan) along with a reduction of temporal resolution of the lateral scan, the 4 orders of magnitude faster shockwaves induced by a short treatment pulse in the sub 10 μs scale cannot be directly resolved in an 2D image. This becomes even more restricting in typically slower ultra-high-resolution systems with a spectral bandwidth of more than 100 nm that supports an in vivo axial resolution of about 3 μm or below, which is required for clear visualization of cellular structures in the RPE layer and differentiation from its surrounding tissue .
11.3 OCT for SRT Dosimetry
11.3.1 Hypothesis of Fringe Washouts in M-Scan OCT
The origin of the above-mentioned complete OCT signal loss can be explained by coherent signal washout. The coherent decorrelation or “fringe-washout” results from rapid fluctuations of the spectral phase, which are faster than the integration time of the detector, such that the time-averaged signal becomes almost zero. This effect is well known in OCT images of larger blood vessels, especially for those oriented parallel to the measurement beam. Their lumen appears dark due to the fast turbulent blood flow of moving cells that generate fast modulation of the phase . This effect differs from intensity decorrelation, where the variation between successive OCT scans is numerically calculated for contrasting motion, while in coherent decorrelation the OCT signal’s phase is already destroyed prior digitization. In coherent or phase decorrelation, the tissue scatterers are moved by a fast oscillating sound field at amplitudes in the range of the axial resolution, therefore the spectral phase is altered quicker than the exposure time of the irradiation pulses.
This ability to discriminate effects due to sole heat-related changes from microbubble-related ones is a viable advantage to simple reflectometry. The reflectometry utilizes a swelling signal that is supposed to correlate with vapour-bubble size . In OCT, however, no growing of signal intensity is found throughout the M-scan within a burst sequence, which indicates that decorrelation superposes the increase in reflectivity and can act as a sensitive indicator for bubble build-up. The distortions introduced are not completely independent of the bubble size; although there seems to be a slight upward drift within and between successive scans of the inner limiting membrane (ILM) signal edge that might indicate a vertical shift due to bubble formation within a burst-sequence; this also could be contributed to thermal expansion and is hard to interpret due to the speckle. In contrast, the length of the black-outs correlate well with radiant exposure/fluence and stronger pulses also experience a stronger structural change as well as an elongated relaxation period (Fig. 11.9).
11.3.2 First Pre-clinical and Clinical Studies
11.3.3 Future Developments Towards Reliably Detecting the Microbubble Threshold with OCT
Several strategies for limiting the exposure beneath the threshold for retinal heat damage have been discussed and exploited. Simply keeping the average power well beneath the average levels for coagulation misses the therapeutic window that spans between pure RPE and extended retinal damage due to the strong local absorption variations of about an order of magnitude. With typical average excursions of ~0.5° during microsaccades that mostly stay smaller than 1° corresponding to shifts of ~288 μm on the retina during fixation focussing of treatment spots to ~200 μm is challenging. Therefore, repetitive application of short pulse bursts between microsaccades (~200 ms) to shorten the time frame below the instability of the human eye during fixation, is one way to adapt to the situation as it was suggested by the reflectometry controlled SRT model [89, 92].
Under the presumption that the appearance of a microbubble already indicates irreversible damage to the RPE and OCT reliably relates detected shockwaves to microbubble appearance, another approach is more sensible. Here the pulse train is successively ramped up in fluence until a MBF signal is reached or the maximum permitted exposure at full absorption is reached (see also Fig. 11.7a).
To successfully utilize a ramp mode, adaptive laser pulse energy control together with a fast and reliable method for exposure interruption is required. On top of the signal washout, which can be detected by simple signal thresholding the high sensitivity of OCT carries more information hidden in the sub-structure of the M-scan, including thermally induced fluctuations and possible secondary changes due to the microbubble formation. One way is to use human interpretation of the signals and to train a synthetic classification network. The network successively learns how to interpret different preselected signal features such as artifacts that compare sub-blocks with and without laser exposure, speckle changes, changes in the speckle variance or spectrum, similar to the TF analysis mentioned above and associates their appearance to the outcome. Besides the teaching with less clear in vivo data and direct human observation of the OCT signals together with TF analysis, this network can also be trained with ex vivo data that is based on histological results to objectively relate to selective RPE damage. As assumed, features found within the first 100 ms after exposure deliver the highest yield, while later portions of the signal between 100 and 300 ms after the exposure pulse continuously lose their relation to the distortion. Such automated analysis boosts prediction to success rates from initial ~60% beyond 90% when targeting for 95% specificity .
Artificial convolutional neuronal networks can take this approach even a step further to select their own internal features, rather than predefined ones. This currently en vogue industry driven approach has the advantage of exploring and utilizing information that has not been considered by the human supervisor, but also bears the risk to train features that are not directly related in a cause-effect relationship that is further encoded in deeper hidden network nodes. A typical example is an artificial neuronal network that is trained to distinguish dogs from wolves, but in the end only learns to react on the snowy background most frequently found in wolf pictures. Similar risks are found in the SRT application where patient, image background signals or system properties can influence the results, unless very large numbers of verified examples are used. The training of such a network on OCT signals, however, also brings novel insights that are significantly harder to quantify by a human observer. When investigating the significance of signal components during the time course for predicting the accuracy of the treatment outcome, it can be shown that window sizes of ~4 ms which also contain information from the status 200 μs prior to the applied pulse for the utilized system achieve the highest precision and recall, or in other words the lowest false positive and false negative rate . An explanation for this behaviour is that 200 μs before exposure, relatively slow changes due to eye-motion do not affect the baseline, while longer ones already exhibit distortions.
For integration into a clinical system these artificial neuronal networks have the advantage of shifting the complexity into a hidden layer that requires enough high-quality data for training, but thereafter can be realized with hardware that is capable to perform decisions in real time even with these sophisticated tasks. Specialized hardware with low energy consumption, high efficacy and speed is currently becoming available. Nevertheless, a black box system for application in laser therapy implies strict control mechanisms. Assuming 5–30 steps of treatment pulses with increasing energy within the stable exposure window of 200 ms, reaction times of ~3–20 ms are the target for ramp interruption. Considering electronic acquisition and laser driver delays, such a decision has to be made within a millisecond or has to be performed one or several pulses later when the pulse energy has already grown. Optimizing the ramps start, end levels and slope will be a question to be answered together with the choice of the definite detection mechanism.
The following chapter goes back from this outlook to the current technical efforts and improvements. It describes in detail the realization of a complete prototype setup combining OCT-based dosing control, therapy planning and documentation.
11.4 SRT Module Integration into the OCT Platform
Early-on the Heidelberg SPECTRALIS platform, originally built on a confocal scanning ophthalmoscope already hosted a robust spectrometer-based frequency-domain OCT system and was capable of tracking lateral eye motion, circumventing the above-mentioned issues by establishing a different approach: Here the eye tracker allows for sufficiently precise and fast online tracking of the scanners lateral position such that imaging of OCT cross-sections become re-entrant. This means that scans can be interrupted and restarted at any point during the scan, and that imaging sessions can be repeated in long-term studies as “follow-ups”, since the tracking algorithm aligns the optics with the long-term stable retinal vasculature, even with poor quality ocular media. Furthermore, the modular design of this commercial multimodal platform also provides access to other wavelength ranges, which can equally benefit from the eye tracker, as well as the integration of other optical measurement or treatment devices.
In a first collaboration with Heidelberg Engineering, the optoLab of the Bern University of Applied Sciences in Biel has modified the mechanical setup of the SPECTRALIS platform and has successfully integrated a second OCT spectrum-scanning light source at 1050 nm into the system. With this setup named SPECTRALIS Hydra, a simultaneous acquisition of OCT images in both wavelength ranges became possible. First clinical studies at highly myopic children have been performed in Hong Kong to examine early aged myopia .
Learnings from this implementation including the mechanical setup and electronic communication enhancements were the basis for integrating a retinal treatment laser instead of a second OCT light source. This new experimental setup named SPECTRALIS Centaurus offers a fast, smart and minimally invasive clinical tool for performing SRT.
Overview of parameters used for different SRT investigation systems
Porcine ex vivo
Human in vivo
Porcine ex vivo
Porcine ex vivo
OCT scan speed
With first encouraging ex vivo experiments performed at the optoLab and the Medical Laser Center Lübeck (MLL) as well as the technical safety evaluation, it is planned to validate the SPECTRALIS Centaurus system for clinical in vivo studies.
11.5 Conclusions and Outlook
The SPECTRALIS Centaurus system integrates a μs pulsed treatment with a high-speed confocal laser scanner in a compact design. Despite not being a commercial device, the platform has shown a possible way to clinical implementation. In addition to the long-term stable opto-mechatronic design, the “commercial-level class” platform features high reliability, low maintenance, combined with a simple and user-friendly interface and remote maintenance of the device. The integration of a fast, high-resolution 3D imaging modality with therapeutic capabilities enables detailed planning of precise control during potentially harmful treatment. The exact positioning of the treatment laser pulse in time and space, independent of the patient’s movement, as well as the possibility of immediate pre-/post-diagnosis, and a feedback system allows for an unrivalled accuracy and fulfilment of the physician’s pre-set targets including safety features that avoid overexposure of the retina. Current approaches for limiting exposure are already well developed and have started to become indispensable tools for minimally invasive optical surgery. The ophthalmic instrument market has shown that contactless, all-optical technologies have always been favoured by clinicians and patients alike, due to their simpler, speedier and safer application with higher convenience and lower risk of mechanical injury or of cross-contamination. Even though the principle mechanisms of laser tissue interaction are well known, the variability of biological materials and delicate alignment of the treatment system demand for considerably increased control for each individual exposure. Precision-guided in vitro experiments already suffer from variations caused by inhomogeneity of melanosome density and absorption coefficients within the RPE layer. To transfer the results of ex vivo trials to clinical ones that add multiple layers of complexity and variation, a high reproducibility of the instrument during clinical application is required. Literally in focus of the research, the critical laser parameters are linked to the available technology and reveal only parts of the theoretical possibilities. As novel compact and cost-effective lasers open up new possibilities, the perfect choice of laser treatment for certain diseases will require more advanced studies that will definitely benefit from a multi-functional device that provides full control during the application of the parameters, along with an advanced diagnostic toolset for planning, performing and reviewing or even aftercare. This not only provides the surgeon with a more precise instrument that improves the statistics, but also allows the patient to benefit from an optimal and individual treatment that results in a therapeutic outcome which close resembles the original plan and mitigates risks and unwanted side effects.
We gratefully want to acknowledge the continuous and valuable support from the team members from Bern University of Applied Sciences in Biel: Christian Burri, Michael Peyer, Patrick Steiner, Daniel Kaufmann, Mathias Mooser, Patrik Arnold, Tiziano Ronchetti, Volker M. Koch, Joern Justiz, Anke Bossen, Patrick Morgenthaler, Dominik Inniger, Christoph Meier, the Medical Laser Center Luebeck: Dirk Theisen-Kunde, Veit Danicke, Alessa Hutfilz, the colleagues from Meridian AG, Thun: Michael Stetter, Rudolf von Niederhäusern, Eric Odenheimer and the Heidelberg Engineering team: Stefan Schmidt, Michael Reutter, Joerg Fischer, Tilman Otto that enabled the development and implementation of this technology.
- 10.Marshall J. Thermal and mechanical mechanisms in laser damage to the retina. Invest Ophthalmol Vis Sci. 1970;9(2):97–115.Google Scholar
- 14.Marmor MF, Wolfensberger TJ. The retinal pigment epithelium. Oxford: Oxford University Press; 1998.Google Scholar
- 15.Gabel VP, Birngruber R, Hillenkamp F. Visible and near infrared light absorption in pigment epithelium and choroid. Congress Series: XXIII Concilium Ophthalmologicum. 1978;450:658–62.Google Scholar
- 18.Gabel VP. Die Lichtabsorption am Augenhintergrund. Habilitation. Munich: Ludwig Maximilians-Universität München; 1976.Google Scholar
- 31.Fritz A, et al. Dynamics and detection of laser induced micro bubbles in the retinal pigment epithelium (RPE). Proc SPIE-OSA. 2007;6632:1C1–11.Google Scholar
- 32.Fritz A, et al. Dynamics of laser induced transient micro bubble clusters. Proc SPIE-OSA. 2009;7373:0D1–9.Google Scholar
- 47.Brinkmann R, Roider J, Birngruber R. Selective retina therapy (SRT): a review on methods, techniques, preclinical, and first clinical results. Bull Soc Belge Ophtalmol. 2006;302:51–69.Google Scholar
- 49.Boettner E, Wolter J. Transmission of the ocular media. Investig Ophthalmol. 1962;1(6):776–83.Google Scholar
- 55.Roider J, et al. Clinical results after subthreshold (RPE) photocoagulation in macular diseases. Biomedical Optics, OSA Technical Digest. 1999:11–13.Google Scholar
- 65.Roider J, Brinkmann R, Birngruber R. Selective retinal pigment epithelium laser treatment - theoretical and clinical aspects. In: Fankhauser F, Kwasniewska S, editors. Lasers in ophthalmology - basic, diagnostic and surgical aspects. The Hague: Kugler Publications; 2003. p. 119–29.Google Scholar
- 66.Roider J, et al. Therapeutical bandwidth or selective RPE-photocoagulation treatment by repetitive mu s-laser pulses (527 nm) - first clinical results. Invest Ophthalmol Vis Sci. 1997;38(4):404.Google Scholar
- 68.Brinkmann R, Schüle G. Method and device for non-invasive regulation of temperature with radiation, especially laser radiation of biological tissue. Patent EP 1 279 385 B1, US 6,830,567; 2002.Google Scholar
- 70.Seifert E, et al. Automatic irradiation control by an optical feedback technique for selective retina treatment (SRT) in a rabbit model. Proc OSA-SPIE. 2013;8803:031–6.Google Scholar
- 77.Seifert E, et al. Selective retina treatment (SRT): towards an acoustically controlled automatic dosing. In preparation.Google Scholar
- 81.Hasgall P, Di Gennaro F, Baumgartner C, Neufeld E, Lloyd B, Gosselin M, et al. IT’IS Database for thermal and electromagnetic parameters of biological tissues; Database Tissue Prop. [Internet] 2018; Available from: https://itis.swiss/database.
- 86.Steiner P. Ultra-high resolution optical coherence tomography retinal tissue monitoring for dosimetry control during selective retina therapy laser treatment. PhD thesis. Bern: Bern University; 2015.Google Scholar
- 89.Roh YJ, et al. Selective retina treatment (SRT) automatically controlled by a real-time reflectometry in a rabbit model. Invest Ophthalmol Vis Sci. 2013;54(15):4136.Google Scholar
- 91.Kaufmann D. Combining SRT with OCT for automated dose-control [Master Thesis]. Biel, Bern: Universität Bern und Berner Fachhochschule; 2016.Google Scholar
- 95.Mooser M, et al. Quantitative, simultaneous & collinear eye-tracked, high dynamic range optical coherence tomography at 850 and 1060 nm. In: European Conference on Biomedical Optics. Optical Society of America; 2017.Google Scholar
- 96.Burri C. Optical coherence tomography controlled selective retina therapy for in-vivo treatment [Master Thesis]. Biel, Bern: Bern University of Applied Sciences; 2018.Google Scholar
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.