System Implementation of a CMOS vision chip for visual recovery

Transcription

1 System Implementation of a CMOS vision chip for visual recovery Akihiro Uehara a, David C. Ng, Tetsuo Furumiya, Keiichi Isakari, Keiichiro Kagawa, Takashi Tokuda, Jun Ohta, Masahiro Nunoshita Nara Institute of Science and Technology, Takayama, Ikoma, Nara, JAPAN ABSTRACT We have developed a CMOS vision chip, an image sensor with pixel-level signal processing, to replace photoreceptor cells in the retina. In this paper, we describe a pixel-level signal processing, which is to control on the stimulus waveform and the amount of the electrical injection charge. Our CMOS vision chip is an array of a pixel, which consists of a photo detector, a pulse shaper, and a current stimulus circuit. The photo detector circuit generates a pulse frequency modulated (PFM) pulse, which frequency is proportional to the intensity of the incoming light [1][2]. The PFM photo detector is also modified to restrict the maximum frequency of PFM pulse signal for safety neural stimulation [3]. The PFM pulse signal should be converted into suitable waveform for efficient neural stimulation. We have employed a pulse shaper to generate one stimulus pulse from one PFM pulse. The pulse parameters (i.e., pulse duration, polarity, etc) of the output pulse signal are controlled by the external signal. For the electrical neural stimulus, the stimulus intensity is given by the amount of the electrical injection charge. The amount of the injection charge should be enough to evoke a phosphene but should be low to avoid the damage of the retinal tissue caused by the excess charge injection. In our prototyped CMOS vision chip, the stimulus current amplitude is used to control the amount of charge. The 6-bit binary-weighted digital-to-analog converter (DAC) with 2µA resolution is used to control the stimulus current amplitude. Keywords: Retinal prosthesis device, CMOS vision chip, pulse frequency modulation 1. INTRODUCTION The feasibility of visual prostheses by electrical stimulation has aroused a great interest. Thousands of people are suffered from retina degeneration for retinitis pigmentosa (RP), and age-related macular degeneration (AMD) [4]. In the cases, photoreceptors are degenerated, however the rest of the neural tissues of the retina are still alive. To recover the vision, visual prosthesis by electrical stimulation of the remaining retinal layer has been discussed [5][6]. Retinal prosthesis devices are classified in the viewpoint of implantation place: epiretinal and subretinal. In the epiretinal approach, a device is placed in contact with the nerve fiber layer, while in the subretinal approach, a device is placed underneath the retina [7]. Our approach is the development of electrical devices that could be implanted in the subretinal space as shown in Fig. 1 [8]. The reason why we introduce a subretinal approach is as follow: The device stimulates the early stage of the visual system so that potentially high resolution input to the outer retina could be achieved. A simple photodiode array has been used as an implantable device in the subretinal space mainly due to its simple configuration without power supply [8]. The photo current is directly injected into the retinal neurons. However, due to poor photosensitivity, photo current is not large enough to excite the retinal neurons. We have proposed and fabricated the PFM vision chip. In our prototyped PFM vision chip, a photodiode is used with integration mode to gain high photosensitivity. A pixel consists of a photo detector, a signal processor, a current stimulus circuit and a stimulus electrode. A Frequency of the photo detector is proportionate to the intensity of a Permanent address: Nidek Co.,Ltd., 73-1 Hama, Gamagori, Aichi, JAPAN

2 incoming light. The control signal restricts the range of the frequency, which is desired for the stimulation of the retinal neuron. The pulse shaper circuit generates the electric pulses with a suitable waveform for stimulation. The stimulus current amplitude is selected according to the 6-bit current register of a pixel. The current stimulus circuit is injecting and sinking current source to synthesize biphasic stimulus current. Implantable Vision Chip in Subretinal Space Implantable Vision Chip Objects Eyeball Amacrine Cells Pigment Epithelium Horizontal Cells Power Supply (Light or Electromagnetic Induction) Retina Ganglion Cells Bipolar Cells Photoreceptor Cells 0.3~0.4mm Figure 1: Concept drawing of implanted vision chip in subretinal space 2. PROTOTYPPED PULSE FREQUENCY MODULATION PIXEL 2.1 PIXEL BLOCK DIAGRAM The block diagram of a pixel is shown in Fig. 2. A pixel consists of a pulse generator, a pulse shaper, and a current stimulus circuit with DAC. To effectively stimulate the retinal cells, controllability of stimulus pulse parameters such as anodic and cathodic pulse duration and the amplitude, and interphase delay is required. In our approach, pulse duration is fixed. The pulse amplitude is used to control amount of the injection charge per pulse. The stimulus pulse frequency is the function of the incident light intensity. RETINA A PIXEL Stimulus pulse timing External power supply Electric Stimulation Current driver Pulse shaper DAC Pulse generator Electrode Memory Retinal Neurons Light from the Outside Figure 2: Block diagram of a pixel

3 Pulse shaping circuit is triggered by the PFM output, this circuits puts out an arbitrary waveform, which is common among the pixels. Once PFM output triggers the pulse shaping circuit, the pulse shaper generates one stimulus pulse. The current stimulus circuit injects a stimulus current into the retinal tissue which waveform is given by the pulse shaper. Fig. 3 shows the prototyped pixel circuit. A pixel has a size 850 µm 300 µm using 0.6 µm 2P3M CMOS process. Figure 3: Microphotograph of the prototyped PFM pixel 100µm 2.2 PFM WITH LIMITED BANDWIDTH OF OUTPUT PULSE FREQUENCY The PFM vision chip is an array of the PFM photo sensor. The PFM photo sensor utilizes representation of output pulses, in which analog intensity of the incident light is converted into pulse frequency. Fig. 4 shows schematic diagram of a PFM photo sensor with a short inverter chain that forms an oscillator with self-resetting. The PFM photo sensor operates like a ring oscillator. The photodiode (PD) acts as a variable current source whose intensity depends on the incoming light, thus the output frequency changes. Vp M pleak VDD 6/0.6 φ1 φ2 Others: 1/0.6(nMOS) 3/0.6(pMOS) Tr. Size =W/L. Vn 0.8/80 M nleak 0.8/26 2/0.6(n) 6/0.6(p) ST 17.5µm sq. (PD) 1µm sq. (Aperture) S 1 S 2 C 1=350fF 2/0.6(n) 6/0.6(p) INV1 INV2 C 2=350fF 7-Bit Binary PFM_OUT counter I[1:7] I[0] Q PFM_OUT φ1 φ2 VSS Period T Sel[2:0] Figure 4: Schematic diagram of the PFM photo detection circuit

4 PFM pulses of a high frequency cause an exceeding charge injection. Charge injection over the threshold is not desirable, because exceeding charge injection could damage the retinal neurons. PFM pulses of a low frequency could cause unsuitable flicker, thus unusual low frequency is not desirable. The frequency of PFM pulses should be restricted within a certain range so that suitable amount of charge is injected into the retinal neurons. To control the characteristics between the light intensity and output pulse frequency, we modified PFM photo sensor. We applied a switched capacitor methodology to a conventional PFM photo sensor as shown in Fig. 4. A series of switched capacitor is inserted between the Schmitt trigger and the inverter chain. In this configuration, the output voltage changes only when S2 turns on. Thus, the upper limit of PFM pulses is given by the frequency of timing pulse 2. M nleak as a current sink forces the cathode voltage of the PD to discharge, so that the lower frequency limit is determined. M pleak as a current source is used to compensate the dark current of the PD. The upper and lower limits of the pulse frequency range are denoted by I nleak f min, (1) ( CPDVth + td I R ) f 1 C1 max. 2RscC = (2) SC 2C2T Here C 1 and C 2 are capacitors of the switched capacitor filter shown in Fig. 4, and R sc and C sc are equivalent resistance and capacitance of the switched capacitor filter. Introduction of the switched capacitor filter enables us to determine arbitrary delay of the feedback loop td=r sc C sc independently of a fabrication process. Light adaptation is the necessary function for the retinal prosthesis device. In other words, photo sensitivity of a pixel should be controllable. Frequency divider with division ratio (2 power by n) is implemented by 7-bit binary counter and 7-input multiplexer. 2.3 STIMULUS PULSE SYNCHRONIZATION The PFM signal is an asynchronous pulse signal. If the stimulus pulse current is asynchronous, extraordinary large power dissipation may occur when all pixels output stimulus current. To avoid this large power dissipation, the stimulus pulse timing should be synchronized. The circuit diagram of the synchronous stimulus pulse shaper is shown in Fig.5 and timing chart is shown in Fig. 6. When the signal PosOut is positive, the current stimulus circuit outputs anodic current into the retinal tissue, and when the signal NegOut is positive, it outputs cathodic current. PFM output signal PFM_OUT is latched in every rising edge of the clock. By the detection of the rising edge of PFM_OUT, the pulse sharper circuit gets into the output enable state (t1 in Fig. 6). When the pulse shaper is in the output enable state, the input signals, PosTim and NegTim, are output to PosOut and NegOut respectively. (t2 in Fig. 6). The pulse shaper gets into the initial state when CLR is positive (t3 in Fig. 6). Thus, the rising edge of PFM_OUT causes to output one stimulus pulse. PFM_OUT D Q C /Q D C Q /Q 0 D Q PosOut CLK CLR C /Q NegOut PosTim NegTim Figure 5: Schematic diagram of pulse shaper

5 t1 t2 t3 t4 t5 t6 t7 CLK PFM_OUT CLR PosTim NegTim PosOut NegOut Figure 6: Timing diagram of the pulse shaper 2.5 CURRENT STIMULUS CIRCUIT The current stimulus circuit is illustrated in Fig. 7. The stimulator is based on a 6-bit binary-weighted DAC which consists of 6 parallel current sources. Each pixel has a 6-bit register file which corresponding bit switches the associated current source. Thus, the stimulus current amplitude is configured through the register file by the external control circuit. The pulse shaper output signal, PosOut and NegOut, controls the stimulus current sourcing and sinking timing. If the reference electrode is connected to GND, dual supply is required to generate biphasic stimulus current. The voltage of the reference electrode is kept to one half of the supply voltage so that we can create a biphasic current pulse from a single voltage supply. Pixel current driver 120/4 50/4 DAC data 6 Bias current 64/6 32/6 2/6 10/0.6 PosOut NegOut Stimulus Electrode 120/4 50/4 Figure 7: Schematic diagram of the current stimulus circuit

6 3. MESUREMENT RESULTS 3.1 LIGHT ADAPTATION OF THE PFM PIXEL The experimental result of output frequency vs. incoming light intensity of PFM pixel is shown in Fig. 8. The highest frequency and the lowest frequency are restricted to 150Hz and 0.5Hz, respectively. Function of wide range light adaptation is achieved by dividing the PFM pulse with adequate division ratio Output frequency [Hz] n=1 n=4 n= Light intensity [lux] Figure 8: Experimental result of input light intensity vs. output pulse frequency. The experimental result of light adaptation of PFM pixel is shown in Fig. 9. The control circuit can configure the largest and smallest light intensity by setting adequate division ratio to the PFM pixel. Because the maximum division ratio of the prototyped pixel is 128, the range of light adaptation is about 20dB Light intensity [lux] fout = 100Hz fout = 10Hz Division factor Figure 9: PFM pixel characteristics of light adaptation.

7 3.2 CURRENT STIMULATION CIRCUIT DAC characteristic of a stimulus current circuit is shown in Fig. 10. The load impedance is 10kΩ and the maximum swing current is set to 200µA. The mismatch of sourcing and sinking stimulus current is less than 1% of the maximum swing current. Current Amplitude (ua) uA UP 200uA DN bit Data Figure 10: DAC characteristics. The load impedance is 10kΩ. 4. CONCLUSIONS We have developed a CMOS vision chip, an image sensor with pixel-level signal processing, to serve as a replacement for degenerated photoreceptor cells in the retina. Our aim is to develop a CMOS vision chip which is implanted under the sub-retinal space. Instead of the degenerated photorecepter, each pixel of the implanted CMOS vision chip detects incident light and injects electrical charge into the retinal tissue to evoke a number of phosphenes. To effectively stimulate the retinal cells, controllability of stimulus pulse parameters such as anodic and cathodic pulse duration and the amplitude, and interphase delay is required. In our approach, pulse duration is fixed. The pulse amplitude is used to control amount of the injected charge per pulse. The stimulus pulse frequency is the function of the incident light intensity. For that purpose, a pixel consists of a pulse generator, a pulse shaper, and a current stimulus circuit with DAC. The pulse generator utilizes representation of output pulses, in which analog intensity of the incident light is converted into pulse frequency. The frequency of PFM pulses should be restricted within a certain range so that suitable amount of charge is injected into the retinal neurons. Frequency divider with division ratio (2 power by n) is used to realize the light adaptation function. If the stimulus pulse current is asynchronous, extraordinary large power dissipation may occur when all pixels output stimulus current. To avoid this large power dissipation, the pulse shaper generates synchronized stimulus pulse. The stimulator is based on a 6-bit binary-weighted DAC which consists of 6 parallel current sources. The mismatch of sourcing and sinking stimulus current is less than 1% of the maximum swing current. ACKNOWLEDGEMENTS This work is supported by NEDO (New Energy Development Organization) in Japan. REFERENCE

8 1. J.Ohta et al., Jpn.J.Appl.Phys., 41, 2322, X.Q.Liu et al., Proc. of SPIE, 4306, Tetsuo Furumiya et al., Proc. of SPIE,4829, 969, J.G.HollyField et al., Retinal degenerative diseases and experimental therapy, Kluwer Academic/Plenum Publishers, E.Zrenner et al., Vision Res., 39, 2555, Alfred Stett et al., Vision Res., 40, 1785, Alan Y.Chow et al., IEEE Trans. Neural System and Rehabilitation Eng., 9, 86, E.Zrenner et al., Ophthalmic Res., 29, 269, phone ; fax ; Nara Institute of Science and Technology, Graduate School of Materials Science, Takayama-cho , Ikoma, Nara , JAPAN