Neurobionics group research

Positions  According to the Royal National Institute for the Blind (RNIB), there are 2 million people with visual impairment in the UK, of which 180,922 were registered as partially sighted in 2005. The most common cause of impairment is due to photoreceptor degenerations (48.5%) followed by Glaucoma (11.5%), diabetic retinopathy (3.4%). While hereditary diseases such as the Retinitis Pigmentosa (RP) group remain fixed, the increasingly ageing population and expanding waistlines are enhancing the prevalence of aged related macula degenerations and diabetic retinopathies respectively. Furthermore the rate of retinopathies in premature babies is around 16%. The (American) National Federation for the Blind estimate that it costs the state £600,000 in a lifetime of support and unpaid taxes for one blind person. The social cost to the individual is of course priceless.

While refractive errors can be corrected with glasses and cataracts via surgery, degeneration of the photoreceptor cells is much less treatable. There is much research into the exact mechanism causing these conditions, and some of the genetic defects causing RP are now being better understood. Additionally the vascular defects which cause macular degenerations such as age related macular degeneration (AMD) and diabetic retinopathy are also becoming better known. However there is very little treatment for these conditions at the present time. What treatment there is such as anti-VEGF drugs (for wet AMD) and photodynamic therapy only slow progression, but cannot return lost vision. Thus other researchers are presently engaged in genetic therapies, tissue transplantation, and pharmaceutical intervention.

In my own case I am interested in two important fields of research:

Augmented Vision: is a method whereby we maximize the information throughput from the eye to the visual cortex by pre-filtering the visual scene and feeding this back to the patient through virtual reality headwear

Optoelectronic Visual Prosthesis: For individuals whose sight has deteriorated to the extent that there is no longer any functional vision, we are investigating a revolutionary form of optoelectronic prosthesis method for returning vision.

Forms of Retinal Degeneration:

Rod and cone dystrophies:  retinal dystrophies, often classed as retinitis pigmentosa are characterized by the inactivation and loss of photoreceptors in the retina. The initial symptoms depend on the nature of the degeneracy. The loss of rods leads to poor night vision, significant loss of the periphery (Tunnel vision) and thus motion sensing. Loss of the cones lead to severe reduction in visual clarity, photobleaching effects at high light sensitivity, and a loss of colour contrast. Often, secondary degeneration of the retina vascular structures can lead to a degeneration of other photoreceptors and eventually to total blindness.

In the earlier stages, augmented vision systems can help improve interpretation of the visual scene. In later stages, while the light sensing photoreceptors are destroyed or inactive, the bipolar and retinal ganglion cells are still intact, thus there is applicability for retinal prosthesis.

Macula degenerations: Dry AMD results from an inefficiency of the retinal vasculature to remove waste products, resulting in druzen. The druzen can initiate photoreceptor cell death causing localized loss in visual acuity there is no effective treatment or cure. More serious is when the body responds by developing new blood vessels which leak causing massive localized cell death. Diabetes has a similar effect, causing diabetic retinopathy. In both these cases the blood mediated cell death can cause a high degree of mortality of retinal ganglion cells. Traditionally the only treatment was through laser therapies, but the development of anti-VEGF drugs have improved things greatly

In these macula degenerations there is usually some remaining functional vision, and thus the scope for the first generations of retinal prosthesis will be limited. Thus augmented sensory systems could provide the greatest benefits at this stage (in conjunction with drug therapies).

Augmented Vision systems

Retinal dystrophies can lead to the total loss of vision in their final stages. However, in the early and intermediate stages there is some remaining functional vision. The visual cortex performs impressive image processing to use this residual trickle of visual information to build up the visual scene, but is hampered by having to (literally) guess the missing image components.

Thus enhancement of the visual scene to take advantage of the remaining feature recognition capabilities of the degenerate retina can greatly improve visual recognition. Head up displays have been used in aviation for many years to provide basic information overlaid on the visual scene. These have been bulky power hungry systems not suitable for daily life. Traditional image processing algorithms are carried out using sequential processing. To achieve video rates, millions of pixels must be rendered in milliseconds, requiring high power processors running at Ghz frequencies. Thus, both the visual feedback devices and power efficient image processing systems need development.

We have been developing a functional model of the degenerate retina by performing patient trials in the John Radcliffe Hospital. We aim to develop a set of image enhancement algorithms which are most effective at improving visual recognition. We are also using our knowledge of the incredibly efficient parallel processing structures of the retina to implement our algorithms in highly efficient silicon architectures. At present we use commercial virtual reality headwear but are investigating more compact systems.

Optoelectronic Visual Prosthesis

This project takes a very novel route compared to previous attempts at sub and epi-retinal implants. In those systems, the high power consumption required by the stimulating electrodes have prevented high resolution implementations. Our approach, is to use novel light sensitization agents to allow us to stimulate the retinal ganglion cells with light rather than electricity. In so doing, it is possible to bypass many of the power and biocompatibility issues involved in physically touching the neurons.

Neuron Photosensitization
Presently most prostheses rely on stimulating nerves with pulses of electricity. It is thus very hard to be precise. In the case of retinal implants, it is not possible to target specific information pathways. It is additionally very hard to stimulate the neural ‘song’ which the thalamic and visual cortical regions of the brain are expecting. 

In contrast, my group is using a revolutionary new approach. We have the ability to genetically manipulate individual types of neurons to be sensitive to light. This is carried out with the incorporation light gated ion channels such as channelrhodopsin on the cell surface of the neurons. With this technique it is possible to target individual cells and define individual action potentials. We can tune the neural ‘song’ up to 40Hz which is sufficient for the parvocellular retinal pathway, and are looking at improving kinetics for the phasic retinal ganglion cells of the magnocellular pathway.

I have ongoing collaborations with the Hankins lab at Oxford University and the Burrone Lab at Kings College, London, to develop the biology, and with Dr. Konstantin Nikolic, also at the Institute of Biomedical Engineering, Imperial College, to develop a better understanding of the biophysics of these opsins.

 Optoelectronic retinal stimulator 

Effective optical stimulation of large numbers of neuron cells requires light emissive arrays which can be individually addressable and provide sufficient illumination intensity to stimulate the neurons. In the case of channelRhodopsin, Instantaneous light intensities of ~0.1 W/cm2 are often required, which is much higher that any commercial LCD or OLED array.

We therefore use arrays of novel high power GaN LED’s, which are fabricated by the Dawson group at in the University of Strathclyde and optically functionalised by the Neil group here at Imperial College. To gain maximum benefit we flip chip the light emitters onto unique CMOS control chips which we fabricate in collaboration with the Drakakis group at Imperial College. In this way, we are capable of independently stimulating thousands of addressable points, achieving maximum efficiency.

Retinomorphic imaging systems
Prosthetic visual systems need to carry out processing to mimic that of the retina which they attempt to bypass. The processing takes the form of contrast invariant image segmentation via spatial, chromatic and temporal derivatives of the original image. The exact functioning of the retina can be complex, but the mathematical functions can be treated analytically. As with the augmented vision concept, we can take a brute force approach which would result in poor performance, or a take inspiration from the retina itself. 

My group is developing intelligent image processing algorithms and architectures which will be able to perform these processing algorithms in real time on mobile platforms. I have previously demonstrated the techniques in computer, CMOS chips, FPGA’s, and more recently on mobile phone processor architectures.