Fig. 7.1  Average stimulation current produced by one, two, and three photodiodes connected in series. The diodes were oriented such that the biphasic stimulation pulses were anodal-first. Data taken for 25 Hz, 500 ms pulses with a 50 mm SIROF active electrode coupled to a much larger return electrode

infrared (905 nm) images onto the subretinal array. A single, photovoltaically-driven photodiode can only produce up to 0.6 V at physiologically safe light intensities [35], a fraction of the 1.4 V electrochemically-safe “water window.” By providing a pulsed bias voltage and utilizing the diodes in a photoconductive rather than photovoltaic manner, they can produce bi-phasic currents sufficient for neural stimulation, and limited only by the electrode charge injection capacity [35]. The common photodiode bias is provided by a periocular coil-based system. Recently Palanker’s group proposed the use of series photodiodes to receive sufficient current photovoltaically [34]. The voltage increase afforded by series photodiodes, combined with the nonlinear electrochemical capacitance of iridium oxide electrodes [14], greatly increases the attainable current, as shown in Fig. 7.1. Since pulsed infrared illumination is directly converted into electric currents sufficient for stimulation, there is no need for a wired connection to a separate power-receiving module. The pixels do not even need to be physically connected to each other. The arrays may be separately placed into the subretinal space, greatly simplifying surgery.

The information transfer rate C from goggles with N pixels operating at S levels of gray at frame rate R can be estimated in a manner similar to (7.2). With an XGA LCD display (N = 1,024 × 768) operating at 25 Hz and 128 levels of gray, the data rate is C = 138 Mb/s. The limit in this approach is clearly on the receiving end of the system – the photodiode array and its interface with the retina.

7.2.5  Thermal Safety Considerations

Power losses due to tissue absorption and intrinsic imperfections in the receiving circuitry lead to heating. For coil systems this includes absorption of RF radiation in tissue between the transmitting and receiving coils, resistive losses in the coils

themselves, and losses in the rectifying circuitry. In photodiode systems this includes light absorption in ocular pigments such as melanin and in the implant itself. In both system types, the resulting tissue heating must be understood and controlled to within acceptable safety limits.

Tissue RF-absorption has a strong frequency dependence, increasing exponentially beyond a few MHz [43]. However, power transfer efficiency also increases with frequency due to the linear increase of the quality factor Q. There is an optimal frequency region balancing these counteracting effects where RF tissue exposure is minimized. Most coil designs operate at a frequency between 1 and 10 MHz [19, 33, 67]. Once a frequency is chosen, there exist design methodologies to maximize receiving circuit efficiency [27, 28]. Such systems can have power-transfer efficiencies exceeding 65% [28].

Photodiodes are rather inefficient at converting light into electrical power. A photodiode’s maximum conversion efficiency (ratio of current output to incident light) typically does not exceed 0.6 A/W. Since photodiodes produce a photovoltage of at most 0.5 V at physiologically safe light intensities [35], 1 W of incident light power cannot produce more than 0.3 W of electrical power – an efficiency of at most 30%. Thus, photovoltaic retinal stimulation is a rather energy intensive task, rendering it imperative to examine safety limits for intense retinal illumination.

According to established ocular safety standards [3, 60], the maximum permissible retinal irradiance for prolonged exposure to near-IR light is 2.8 mW/mm2.1 Similar thermal considerations apply to heating of the iris. Peak irradiance can significantly exceed the average value during short pulses if the duty cycle is decreased. For example, in a goggles-based system with 1 ms pulses delivered at 25 Hz, the duty cycle is 1/40. The peak irradiance during the pulse can then be increased by a factor of 40 – to 112 mW/mm2. Assuming a light-to-current conversion efficiency of 0.4 A/W, the maximum current that can be produced by photodiodes with this irradiance is 45 mA/mm2, corresponding to a charge density of 45 mC/mm2. This value exceeds the retinal stimulation threshold on large electrodes by at least three orders of magnitude [24].

Most retinal-heating studies to date have been acute; little data is yet available on the effects of chronic retinal heating. However, it has been observed that chronic lens heating by 2–3°C can lead to cataract formation [57], in which case chronic heating due to electronic implants could also cause cataracts. A “less than 1°C” criterion for implantable devices is codified in EU safety regulation [1] since this level is comparable to natural variations of the body temperature [50]. For a diskshaped heater of diameter D which dissipates power P the maximum temperature rise DT in the adjacent medium is [52]

1 The ED50 level for producing a minimally-visible lesion with near-IR light (l = 810–950 nm) for spot sizes larger than 1.7 mm on the retina and exposure times exceeding 1000s is 56 mW/mm2 [42, 43]. With a safety factor of 20, the maximum permissible exposure is then 2.8 mW/mm2.