 Felly mae'n gweithio'r gweithio yma i'r ddweud o'r Llyfrgell Llyfrgell, yn y 100th anferstru yma o'r Llyfrgell Llyfrgell. Mae'r Robert yn ystod, rwy'n meddwl i'n ddweud y 11 yma sy'n gweithio'r ddweud o'r ddweud o'r ddweud. Rwy'n meddwl i'n meddwl i'ch gweithio'r prospect o ddoddifu'r prysgwysau o'r ddweud o'r ddweud o'r llyfrgell ac o'r adroddau ar gyfer. So, Helen Robert persuaded me that this would be good for, maybe give a review of my career and then I started thinking about the directions I'd taken and really what have been my influences both in terms of people and also in terms of work that had been going on in the field at the time which kind of influenced the directions I'd taken. So, I've decided that my presentation today would focus on a handful of research papers which had been particularly influential on me, a lot of tidal sites and well worn anecdotes which I'll fit in as well. And therefore, as Robert said, boring from the format of BBC's Desert Island Dis, I give you a Desert Island Docks. So, your cast away today is me and I've chosen eight papers which reflect various influences right from my first day as a PhD student. So, I'm going to begin my story at Imperial College in 1981. I just finished a degree in physics. I still had ambitions to be an astronaut as everyone else did a degree in physics at that time. And I'm embarked on a PhD in astronomy and I did it in the Blackit Laboratory which is this building here. And on my first day, my PhD supervisor, Dr Brian Morgan, he gave me a stack of papers and said, read that lot. And these were all to do with a new interferometric technique which had been developed in a few years beforehand called stellar interferometry. And basically it was a method of achieving information making measurements at the diffraction limit of a telescope. So, I'm going to say something more about that through my first paper which was written by the inventor of this technique, Antoine Laverie, who published this paper in 1970 and laid the groundwork for long baseline interferometry in optics. The idea is very simple, which I can condense as follows. If we have a perfect imaging system without any distortion, we have waves from a star entering the aperture of our telescope, then the telescope will produce an image which represents, if it's a point star, it will be a point spread function of the telescope. And of course, the larger the aperture of the telescope, then the narrower the point spread function, the better resolution that you get and the more detail you can get of the objects you're looking at, whether it be galaxies or individual stars and so on. But unfortunately, the atmospheric distortion reduces that spatial resolution considerably. So, the wave front that comes from travels many billions of miles perfectly uniform and then it comes through the atmosphere, a few miles of atmosphere and gets distorted. And that distortion in the wave front then causes the image to be blurred out. And this blurring out is pretty much the same whether you have a small telescope or a large telescope. And thus, the very large telescopes that were being built at that time in the 1980s, we couldn't exploit the potential for getting high resolution images. Labry's idea was to take very short exposure images of stars, which essentially freezes the distortion in the atmosphere, which has a temporal correlation time of a few tens milliseconds. So, if you take an image which lasts a few milliseconds, essentially that distortion is frozen. And what you get, you get an interference occurring from different parts across that wave front, which produces an interference pattern consisting of multiple point spread functions, albeit photon noise dominated images. And these multiple images, which is what we call speckles. And these speckles, I say, have a correlation time of a few tens of milliseconds. So, the idea was to record very short exposure images of interesting stellar objects and then process them in order to measure their sizes and obtain information at the diffraction limit of the telescope. One of the highlights of my PhD was to travel to the Siding Spring Observatory in New South Wales. It's about 500 kilometres from Sydney in the outback, actually a very beautiful area surrounded by kangaroos, woken up by kangaroos outside the dormitory every morning. And I got to use this telescope, which is the largest telescope on the site. It's the four metre Anglo-Australian telescope. This is the telescope inside the dome. It's hard to appreciate just what the scale of this thing is. The actual mirror is right down the bottom here, that's four metres. This is a very big structure. And it was my task to spend the night in the cage at the bottom of the telescope. So, this is the rear end of the telescope. The primary mirror is up here. There's a hole in the mirror here where the light finally comes through. So, you put your detection optics up here. And we were recording images on videotape. So, as the telescope swings about pointing different objects in the sky, someone had to sit inside this cage at the bottom of the telescope to make sure that the video recorder was always level. The telescope tilt over, so that was my job. What are PhD students for? I'd spent all night in there. This was the Australian winter time, so the nights were quite cold and quite long. Occasionally I was allowed out to have a hot cup of coffee and then to go back in the cage. OK, get on with it. It's only about eight hours to go. So, I collected a lot of data this way. It was quite a successful project. And we published quite a few things. Another highlight of my PhD time was to attend a workshop in Oracle, Arizona. So, it was held at a Spanish villa that had been built by the then Countess of Suffolk. And she obviously wanted to live in the Arizona sunshine, but she took a lot of England with her. And right in the middle of the Arizona desert, she created an English garden. And so, beautiful lawns, English trees and flowers. Had to be heavily watered to keep them alive. And she had this beautiful house. And when she died, she gave the estate to the University of Arizona. And that's where this conference was held. The Casabella, the House of Gold, later known as Suffolk House. Actually, this site was later used for the biosphere. I don't know if any of you remember the biosphere. I think in 1991 they locked up a bunch of eight people for two years to simulate life for distant exploration of plants. What it would be like to have a self-sufficient environment. I think they all went crazy in the end, but anyway. Anyway, it was at this workshop that I met the author of my second chosen paper. And he called himself Pete Warden. Pete Warden was someone who had the charisma that filled the room. He was amazing. In fact, he wouldn't stop telling you how amazing he was. But I really like that about him actually. He'd recently completed a PhD in astronomy at the University of Arizona. He joined the US Air Force. He clearly had an eye on getting on to the NASA and astronaut training program, which, as far as I know, he never made it that far. But he did tell everyone at the conference that he intended to become a general by the time he was 40. And nevertheless, he had a very successful career at NASA. He became director of the Ames Research Centre in California. In fact, this is on his retirement from that directorship meeting President Obama in 2015. Anyway, Pete, at this conference, presented something that I was very impressed by. At that point in my PhD project, we never had the capacity to do imaging. We were only doing measurements of the distance measurements. We didn't have the facility to make images. And what he presented was images. So his technique, which was very simple, he got these spectral images, these short exposure images of stars. He then used some kind of computer algorithm to identify the brightest speckles, so the 20 brightest. He then found the centres of those speckles. And then, essentially, he would deconvolve those 20 speckle images with these 20, essentially the locations of those speckles. That deconvolution then produced one big speckle, which was the image. And this technique enables him to produce this image of a gamma orionis, which essentially is a point source, and alpha orionis, which is a supergiant star, which is resolvable with large telescopes. And he was able to just show that alpha orionis was clearly larger. And, of course, this was very impressive. To make it even more impressive, he produced this image. And this is totally false colour. What Pete did is he essentially got his image. He then cut around it to give it a circular hard border because people expect to see a star to have a nice hard border. And he freely admitted that these little features on here were well below the resolution limit of the telescope. But, nevertheless, he released that to the press. A term you do nowadays is it went viral. And it was published in quite a few newspapers. And if you went into the gift shop at any of the leading observatories, you could buy a t-shirt with that image on and a mug. And some of you might know Chris Dainty. And Chris Dainty, who was a professor at Imperial College at the time I did my PhD, he got a poster of that and he pinned it up on his office wall. I should say that the fact that the detail here is well below what could possibly be resolved and that outside boundary is purely artificial, you don't become a general if you let physics get in the way of a good story. Before I go on, let me just say a little bit about alpha orionis. It's also known as Betelgeuse. It's a red supergiant star. It's photosphere, that's its visible sphere, is larger than the orbit of Mars. So it's a very big star. It's also not very dense. It's about ten times more massive than the sun. But it says it's radius is truly enormous. Beyond the photosphere, it also has an extended hydrogen atmosphere which emits light at particular interesting lines. And although the star is comparatively young, only nine million years old, it's very short lived. And it is a star that almost certainly will go supernovae at some point, at any point in fact. And when it does, it will be something that will be visible in the daytime to us on Earth. If you want to find orionis, you don't have to look south in the winter time. The southern sky in the winter is dominated by the Orion constellation and alpha orionis Betelgeuse is the very obviously pink star on the top left. So the workshop I attended was hosted by a local astronomer whose name was Keith Hegey. And he offered me a postdoctoral position at the end of my PhD, which I gladly accepted. And this is me. This is Keith Hegey and another astronomer I was working with, Julian. And it was a fantastic time. It was a very productive time. Publisher of the work did some really interesting things. So I'm just going to highlight a few things. My project involved using what was then the world's largest telescope. This is the multiple mirror telescope, which instead of having a single large mirror had six smaller mirrors. Of several reasons, most obvious reason, it's easier to get six small mirrors up a mountain than one large mirror. Nevertheless, with careful optical design, this mirror had, although it had the light collection power equivalent to a single 4.5 metre telescope, it had the resolving power, the potential resolving power equivalent to a 6.86 metre telescope, well beyond anything that then existed in the 1980s. So it was my job, one of my jobs, to try and exploit that to see if we could achieve something at the diffraction limit of this enormous telescope on the top of Mount Hopkins, about two hours drive south of Tucson. So the way in order to get this telescope to operate as if it really was a 6.8 metre telescope it's important that the light paths from the objects to the detector all coincide within the coherence length, which for the kind of filters we were using meant less than a millimetre. So that was really a big engineering challenge and it's amazing that this thing really did work. So with really quite remarkable robustness in the design of the multiple mirror telescope, we were able to achieve this condition such that we could combine all these light paths to within that coherence length. The fine tuning was done by inserting pairs of prisms into the beam, which we would do interactively. So you do that with pairs of mirrors until you can see fringes, then you do that with all respect to one mirror then you add the lot together to get the telescope operating as a single array. My specific project involved implementing a technique that we called differential speckle interferometry, which we can consider this to be an extension of Pete Warden's idea, except that we would extract the speckle positions from the broadband photospheric speckles. So use the nice broadband to see where the speckles are, and then we would then produce more astronomically interesting speckle patterns at a narrowband filter corresponding to the hydrogen emission line, that is the large cloud of hydrogen surrounding the star. So this is a relatively tight speckle where these is a very dispersed image, and by deconvolving this with this we hope to get back to producing an image of the hydrogen envelope around the star. So this was a pair of examples that have been coded on a video, and so this is a broadband continuum, the narrow emission line. First of all we tested it just on a bright boring point source star. When you do that, when you implement the technique on a distant star, what you get is essentially the point spread function of the telescope. Now that's actually quite boring, you'd think. However, in fact this is the image we obtain for gamma or anus, and this is the computed perfect point spread function, which you see agree very well. And although the image is very boring, that image represented a world record in the resolution achieved for an optical image of any stellar object. And I just wish we'd made more of that at the time. If I'd been Pete Warden, that would have been on newspaper in the world, but we kind of did nothing with it really, apart from just putting it in a paper as an addendum to more interesting images. But that was in the history of telescopes, that was a record that had only been established maybe a handful of times since the first telescopes were ever invented, and it's probably been only beaten maybe once or twice since. So anyway, we didn't make much of it. I would have liked to see it on a t-shirt though. These are the images we acquired using that DSI technique. So this is the image I've essentially just shown of gamma or anus, and this is extended, horrible force of colour I really appreciate. Extended atmosphere, hydrogen atmosphere, around batheleges. And those are some nice results, we looked at a few other stars as well. Okay, well, the technique worked well, and it clearly was laying the foundation of a technique that we now call adaptive optics. And around that time, the US military got really interested in adaptive optics, partly because the possibility of compensating for the atmosphere in space-form lasers to fire through the atmosphere and shoot down missiles and so on. Anyway, Keith Hegey and others were getting offers of large sums of money from the military to fund the research, but I was obviously quite excluded from that because being a non-US citizen, I wasn't able to be funded by that kind of work. So I looked around for other challenges at that time. I didn't want to come back to the UK then, I enjoyed my time in the US. So I looked around and I interviewed for a post-op position at the University of Utah, the Department of Radiology there. And I met this guy, Robert Krugos, head of the Medical Imaging Research Lab there in Utah. And Bob was also one of those people whose charisma can fill a room. And he'd made a bit of money from a technique, from a patent that he developed for a technique known as digital subtraction angiography. So he invented that technique, earned quite a bit of money, and he was prepared to invest it into researching something new. That was really the source of the post-op trial position that I was applying for. I told him that I knew nothing about medical imaging and then I realised that was an advantage as far as Bob was concerned. But he asked me what I did know about and I said, well, I know about optical imaging. He said, good, okay, investigate that then. He said, do that, investigate optical imaging. And he essentially gave me two years to see what I could come up with. It was that free. He said, go on, I'm trusting you to go away, come back and do something with optical imaging and let me know how you get on in two years. So, of course, I began with a literature search. Now I need to choose my third paper. Well, one of the first papers I found that initial time was this published by Ernst Carlson. It appeared in an issue of a magazine called Diagnostic Imaging. And it really said that a technique known as breast transillumination really could solve the whole problem of imaging early breast cancer without using potentially harmful x-rays. And it had some very positive results. And what it showed is images like this is a healthy breast. It involved nothing more than placing an infrared source underneath the breast and then recording an image with an infrared sensitive video camera which is then observed on a VDU. And it showed these images of the healthy breast and then another breast with a large cancer within it. Well, it's exciting. Looks like somebody's doing something in this area. Looks rather positive. I continued my literature search. And then I found this paper. This was published by Barbara Monseys, who's a radiologist of Washington University in St. Louis. And she essentially poo-pooed the whole thing, really. She got hold of one of these same commercial devices that Carlson had used and she conducted her own study and then compared it directly. I did a blinded study and compared it with x-rays and found that the light scanning achieved a specificity of about 86 percent but a sensitivity of only 58 percent. There was a lot of false positives which were actually also acknowledged by Carlson's original paper. I contacted Barbara Monseys who then sent me these images which don't actually appear in her paper but I just asked her what she'd found and she sent me these as an example of really why the technique doesn't work. First of all, this and this are both nipples, so ignore that. This here is a tumour, probably two or three centimetres across inside the breast. This here, whatever it is, is perfectly healthy. And to be honest, they're not very different from each other and this just illustrates why this transillumination method was never going to work. Now, I can't really select either the Carlson paper or the Monseys papers to take to my desert island because they kind of cancel each other out. So I'm going to select another paper which I found during the same time and that was this one. This was called X-ray gogs, preliminary evaluation of a new medical imaging modality. Now, I'm sure some of you may have heard of X-ray gogs. X-ray gogs, all the rage in children's comics in the 1950s and 1960s, when you put them on it enables you to see your bones in your hand, you can see through flesh, really quite remarkable. In fact, scientific marvel of the century. Of course, it's not such garbage as you might think. I have a pair here. You do realise you're all naked at the moment. You can see through clothes as well, which was another big advertising point for the system. Anyway, this was a lovely paper. It published in the April 1 edition of the journal. The guy with the bona fide radiologist who did this study, he was a pediatric radiologist, so he worked with small children. Of course, children love this kind of thing. He noticed that X-ray gogs provided the correct diagnosis in all cases where the conventional study revealed no abnormality, which was 60 patients, so spot-on for 60 patients. So no false positives whatsoever. His conclusion in his paper was because of low costs and lack of ionising radiation, X-ray gogs are recommended in all cases where radiography is not normally required or results of radiographic study will not influence the patient management or the diagnosis has already been established by other names. Anyway, I love this paper and I thought, well, I really decided that transillumination, as it then existed, was unlikely to be better than X-ray gogs. So I needed to think about doing something different. So I started to think about how else could we use light that doesn't just mean not just shining light through the tissue. So I came up with the idea of trying to measure the flight time of photon as it passed through tissue. So I started doing some computer simulations and my idea was if we inject a pulse of light into the tissue, we measure the time it takes for it to fly through, then that time of flight distribution would also tell me something about the spatial spreading within the tissue. So if we have a source of narrow pulses that spreads out, it scatters throughout the tissue and we get this distribution of photons which we now call temporal point spread function. So I did some simulations that show as you use shorter and shorter photons, it then enhances the spatial resolution that you get across the object. So I approached a physicist in the physics department at the University of Utah who had all the kit I needed and he charged me $100 an hour to use it, which this was 1988, I think, which seems like a remarkable amount of money, actually. I feel a bit ripped off now. See it's somewhat excessive. However, I did get free use of his postdoc as well, Dr Cam Wong, and Cam and I worked really well together and I saw Cam only a couple of years ago that came to UCL, visit me. And we embarked on a very productive period. We did a lot of experiments and published quite a few papers. And I also got some large grants and I was offered a permanent position. Just some of the experiments I did. I made this plastic box, transparent box. I put some little plastic transparent targets with some little opaque disks on them at different depths. I filled it essentially with milk. I then shone my laser beam across it, scanned it backwards and forwards. Without the milk, of course, you can see the five targets. With the milk, you can only see the targets near each surface. However, by using time result, by gating the light travelling through and using shorter and shorter time gates, you can then start to see the objects in the centre of the target. So it's a really simple idea and it kind of works, although you soon run out of photons, of course, and it becomes noisier and noisier. And then we had this journal club in Utah, the Medical Imaging Research Laboratory, and I saw this paper, it's my fourth choice, published in 1988 by a group I hadn't heard of. This paper was written by someone called Dave Delpy and this paper, published here in PMB, used exactly the kit that I was using in Utah. This group was measuring the time of flight of photons through tissue, specifically I think it was the head of a rat, wasn't it? I thought, wow, somebody else is as man as me. They did this sort of thing. I was very excited by that. A few years later, in 1991, I attended a conference in Los Angeles where I met Dave and his team. I also was blown away by a particular presentation by Simon Arridge, which described how imaging across diffuse tissues had a robust imaging algorithm for reconstructing the images. This laid the foundation for toast, which is Simon's internationally leading work in this area. Quite rightly, Simon is seen as the father of diffuse optical tomography, the amazing work he's done in the development of toast. Having made this contact with Simon and Dave and this is toast that appears today on the website. It's a remarkable software. I then approached Dave. I think in the summer of 1991, I travelled over to London, visited Dave and I said, Dave, everything I really need to join your group because you're clearly doing everything I'm doing, but I'm really a group of one over in Utah, but you've got a large group. This is clear. I need to join, come here. How can I do that? Dave said, try applying for some fellowships. So I did. I applied to EPSC and Wellcome Trust. Actually, I got both. I was really lucky. I chose the welcome one. It was more money. I came and joined UCL in November 1992. Here I am in the laser lab with my PhD student, David, in the laser lab of Shropshire House. I can't say I have particularly four memories of Shropshire House, but we did some really interesting work. Dave and I also secured a grant from the Wellcome Trust to build a time-resolved imager so we could take the technique or time-resolved imaging into the clinic. This became what we call the monster technique. However, we weren't the first to do this. I've come to my sixth choice, which is a paper by Susan Hintz and David Benarran at Stanford University. David Benarran actually published a paper, I think, in 1993, which described the hardware of his time-resolved imaging instrument. He was interested in, Susan were both pediatricians, were interested in imaging the newborn infant brain. His line of thinking was identical to what we were thinking at that time. He built this system, which involved placing a ring of optical fibres around the circumference of the head. Unfortunately, he had just a single detector. I think he might have had more than one source, but a single detector, you illuminate one source and you can measure one detector. It's nice to get sufficient images to do a tomographic slice. He'd have to move his detector from fibre to fibre. For that reason, his scan times were many hours. Paper doesn't appear to say the number of hours, but I remember in discussions. It was the order of six, seven hours to get a total brain image, which, nevertheless, it was fantastic work. I really admired the whole spirit, the daring of the work. The algorithm was a bit iffy as well. They didn't have toast. They did some basic back projection, which was a bit hard to justify, but, nevertheless, it showed some gross abnormalities on each side of the brain, which were consistent with the existence of hemorrhage. Back at UCL, this is the monster system that we built. Monster was an acronym, but also it was appropriate giving the size of the instrumentation. It was originally meant to be housed in this single rack, but then we found we needed an additional rack. We also needed a cooling system, which is not peering here. Another PC, and this is... That's a real baby, I think, and that's a real nurse. We did some really interesting scans over at the nearest unit at UCLH. My main concentration at that time was to do imaging that was intensity free. The problem with measuring intensity is you've got a problem with coupling. As the light travels into the tissue, the coupling, how much of the light enters the tissue, depends on how hard you press. If you move slightly, the amount of light going in varies. Coupling is a huge problem. It varies all the time, and babies won't necessarily keep absolutely still. So I was very keen on moving away from intensity and attempting an absolute image of the actual properties, and that's what really drove me at that time. This was an example of some successful work on that, but this is with a phantom. This is tissue-simulating phantom. We placed our ring of 32 sources and detectors around the phantom. It contains these rods of different absorbing and scattering properties, and this is the cross-sectional images we expected, and these are the images that Toast generated. We were delighted with this. This was new. No-one else was doing this, and we were quite excited. We then applied it to arm imaging as well, which also worked surprisingly well. This is a couple of MRI images of the forearm of the same subjects we used for these images. These are the scatter and absorption images across the forearm. The scatter images clearly show the bones, and the absorption images were probably dominated by the blood vessels with probably the white spots there, but otherwise it's pretty uniform except where the bones are. We were also going into breast imaging at this time. Adam Gibson was heavily involved in this work, and we published quite a lot of work looking back on applying this method to breast imaging. In this case, our idea was to place the sources and detectors in a ring and then have the, I was going to say victim, the patient lean up with the breast placed within the ring, and then we would inject light into the ring and use the transmitted light to reconstruct an image. We would then reconstruct the properties relative to a reference, which just made the whole process a little easier, make the reconstruction easier. The scan times weren't long, about five minutes per scan, and we got some really good results. This is an example of both the good and the bad of diffuse optical tomography applied to breast imaging. This patient had, so these were the optical images we obtained, right and left breast, and then afterwards, she had an MRI image using a contrast agent. And amazingly, this sort of double appearance of the tumour is also reflected in the MRI image. We thought, wow, this is fantastic. We did this without an MRI magnet just in five minutes, non-invasive, really great, fantastic. But, however, when you look at the healthy breast, you see a feature here quite uncorrelated to anything inside the anatomy. It just might have been a bit of dirt on the surface. It might have been a blood vessel over an optode. And that is the problem with optical imaging applied to the breast. There's too many false positives. So sensitivity, we found to be incredible. Very, very good. We could see everything, but then we could see more than everything. We could see a lot of things which we couldn't really distinguish between what was healthy and what wasn't. By that time, October 2001, Dave and also Tim Mills have built up a really powerful group. Quite a dream team, I would say. You can recognise people that are still around today, of course, more than still around. I think we've got to present a programme grant. It's a Welcome Trust programme grant which enabled us to recruit Nick Everdale, Terence Lung, Adam Gibson. I think Ilyas joined us about the same time as a PhD student. And of course Paul Beard was brought through by Tim Mills. And we had a team of some brilliant talent at that time. And, unfortunately, they've demonstrated that coming through today. OK, I'm going to choose my seventh paper. This was a true game changer. So I'm returning back to the early 1990s. And this paper was among the first, if not the first, to demonstrate the potential of optical imaging as a functional imaging technique. So in this case, what Hoshi and Tamora did, they just placed some optical fibres in a ring around the head and then they observed the changes that occurred in the transmission of light when the brain was given a cognitive task, some mathematical problem to solve. And that produces activation in the relevant parts of the brain. You get this extra blood flow, perhaps extra difference in oxygenation which changes the colour which the light absorbs, which then changes the intensity of light collected at the surface. And this really laid the foundation of DOT as it is today, which is primarily a functional imaging technique, so a true game changer. Suddenly everyone was doing this sort of difference approach, doing functional images, and that included me too. So we went back and we did a whole bunch of different types of imaging to try and see if we could find changes in the brain by collecting two sets of data. So difference imaging is much easier mathematically and provides superior quality images. But it does require change to occur or to be induced. And rather, the force method that we employed was to use a baby on a ventilator and then we changed the ventilator settings to either change the CO2, the inspired CO2 or the oxygen. And we're able to, as you increase the CO2, you get an increase in the blood volume and also the O2 change, obviously change the balance between oxy and deoxy hemoglobin with two different absorption signatures. And we're able to get these difference images, pretty global changes in optical imaging occurring within the brain, possibly showing the ventricles in the centre and some perhaps slightly more localised changes in oxygenation, it's hard to say. So that was quite successful. Then I think Adam Gibson took the lead on this work, which we're very proud of and I've shown this a thousand times. This is just a functional study we did on a series of babies in the hospital where we acquired images with monster during activation, passive activation of the arm. And... That's Adam's voice in the background there, we'll just stop there. And the data that we recorded, we obtained light over the entire 3D brain and were able to reconstruct 3D images of the brain showing the changes occurring in the blood volume within the head and in this particular case we saw a localised change on the left side of the head roughly corresponding to the motor cortex. So these are sagittal slices going from the left ear across to the right ear. So left ear all the way through to the right ear. So this is an increase in blood volume in the left motor cortex, possibly indicating an associated decrease in the frontal region but we couldn't prove that. The sudden interest in doing difference work sort of drove us towards making devices which are a lot easier to use in a hospital. Dave and Nick Everdell designed a system which we now call the NTS, DOT system. It measures purely intensity at two wavelengths but has been really the workhorse of optical imaging here at UCL and we're glad to say elsewhere as well that we've made and solved these systems for other groups around the world and it's been a great success. I never tire of showing this image. This is one of the first images we have acquired, I think, with the NTS system. We've just placed a relatively small probe of optical fibres on the top of a newborn infant as it was going to have a blood sample taken. Once a day a nurse would come along, stick a needle into the baby's heel and extract a small quantity of blood. Of course this hurts and the baby yells and screams but it also produces a pain response in the head and we hoped that we might capture this. We did, in fact the response is enormous and this shows when that counts down to zero that's the point at which the needle was stuck into the baby's head. This is showing change in blood volume measured over the centricortex at the top of the head and we're seeing an increase in blood volume on the left side and a corresponding decrease on the right side and that's because the needle was stuck into the right foot so you get a response on the contralateral side of the head with stealing blood from the other side. What I like about this image is there's no averaging. That's the data. MRI has to average the hell of the data to show an image but with optical imaging, no averaging, that's just the raw data shown as a reconstruction and there is no other technique that could do that. So I come to my final paper and I first saw this excellent work by a group of Joseph Culver at a conference and he just blew us away with the quality of the images that he showed at that conference and I think it brought DOT to a new level, to a new high, really set the standard by which we had to work hard to compete and essentially what they developed was a very high density probe of sources and detectors which they placed over the visual cortex of the head, a sort of bald male volunteers and then they exposed this volunteer to a stimulus of a rotating checkerboard pattern and so the volunteer would focus at a point in the centre of this and so what it's mapping out is a response within the visual cortex to the left and right visual fields and there's this amazing correlation between the position of the stimulus and the parts of the visual cortex which are being stimulated and they did some other experiments as well which were equally impressive and I really think this set a new standard which we had to keep up with. So a couple of years after that we became more ambitious, with Rob Cooper taking the lead on this we were doing simultaneous EEG and optical imaging, whole head optical imaging of newborn infants who are potentially suffering from seizures and we were able to study for a whole hour on a baby that although it was sedated so it didn't move it nevertheless was suffering multiple seizures and this is the EEG data recorded over that one hour and the corresponding raw data so these just intensities recorded between pairs of optical fibres on the baby's head and we see these episodes of electrographic seizures occurring over that one hour period with some corresponding wiggles in our intensity data and I think it was David Holder who helped us interpret this EEG data and we identified six possibly seven seizure-like events over that period. Then what Rob was able to do was reconstruct some images so this is images projected onto the surface of the brain so this is the same surface but there are three different views and there's a red line moving from left to right here showing where we are in the time course and here we have the corresponding changes in blood volume occurring on the cortex of the brain and during each one of these seizures we see this sudden pulsing, it goes red to blue very quickly so we see a sudden increase followed by long slow decrease this has also been seen in single point measurements on the brain but never as a two dimensional image and I think for us this was a bit of a landmark as well in fact we're very pleased with it but it's hard to interpret what it actually means and that really has been the stumbling block at this point I'm finished by just taking us to where we are today Gallab's who built the NTS and have been quite successful in marketing that they've now released a portable modular system known as the LUMO already got a few customers so we're hoping this will be a great success so it also opens up to new areas, new types of potential patients that can be studied with this particular device because it is wearable so you just place as many of these units onto the scalpers you wish each one supporting a number of sources and detectors so as I return to research what I'm currently working on is really to try to get back to trying to get away from just difference imaging all the time I've got some ideas on how we might try absolute imaging that doesn't require needing two states in order to get an image how a single measurement can give you the information you need you don't need an underlying change in order to make an image and that's really what I'm concentrating on at the moment I've got a few ideas about that so coming back to design docs I found the process of collecting papers very interesting and at this point I feel I really should say something profound about the directions a career can take and how it can be influenced by people and by the surrounding research activity that we're exposed to however having thought very hard about this I have to conclude that influence is a pretty random that opportunities are largely serendipitous but if we're very, very lucky we might find something that is both interesting and also potentially useful now if you're stranded on a desert island with a bunch of papers isn't something I particularly relish and albeit eight very good papers it doesn't seem like the most useful thing to have when you're on a desert island but I've given a lot of thought to that too and it had occurred to me had to put them to good use paper boat thank you