WEBVTT 00:00.000 --> 00:11.880 Okay, hello Foster. It's good to be back. My name is Joe, I'm a research for the University 00:11.880 --> 00:18.920 of Glasgow, my second time at Foster, and my talk is about open research about something 00:18.920 --> 00:25.720 called the open flexion microscope, and I've got it from two sort of directions. The research 00:25.720 --> 00:32.720 using the open flexion microscope, we both develop it as a tool, try and improve it, 00:32.720 --> 00:38.920 improve it, increase it's capabilities, but also look at the applications for it, not developing 00:38.920 --> 00:46.920 it as a tool for tools sake, but an entire ecosystem of a journey of everything that is, 00:46.920 --> 00:53.280 in some ways quite a simple device has gone further than I think anyone on the team ever 00:53.280 --> 01:07.120 saw. So, this started with more technical issues. Well, that's a real shame. This started 01:07.120 --> 01:13.760 a few years ago at the University of Cambridge with my boss Richard Beaumont, working 01:13.760 --> 01:21.160 in an optics lab, and for the afternoon board question he was asking himself is, taking one 01:21.160 --> 01:26.840 of the big metal expensive microscopes in the lab, how much of that can you replicate 01:26.840 --> 01:33.080 at home using basic tools, a 3D printer, and off the shelf parts. And the answer 01:33.080 --> 01:39.520 it turns out is quite a lot, otherwise I wouldn't be here. This was the early prototype. 01:39.520 --> 01:45.520 You can see it's a couple of long way since then, but the idea of having three gears, 01:45.520 --> 01:52.960 some plastic parts, and a Raspberry Pi camera, with a basic lens, it's really working 01:52.960 --> 02:00.600 as a digital microscope. So, this was the first published piece of research about the open 02:00.600 --> 02:08.760 flagship microscope, February 2016, so we're currently celebrating 10 years, and it was looking 02:08.760 --> 02:15.120 out, again, basic 3D printers, basic off-the-shelf optical components, and what level of 02:15.120 --> 02:21.440 performance you can get, combining those to make a translation stage, putting a slide on 02:21.440 --> 02:27.360 there and using it as a microscope. It's all based around the idea of flex your hinges, hinges 02:27.360 --> 02:32.280 that rather than parts, swining past it, which 3D printed parts are generally quite bad 02:32.280 --> 02:36.920 out. Instead taking advantage of something that plastic is very good, which is deforming 02:36.920 --> 02:43.920 reliably and repeatedly, the flex your hinges is what gives the stage its translation, 02:43.920 --> 02:50.880 and the fact we are open is what gave open flexure its name 10 years ago. We take this 02:50.880 --> 02:57.000 flex your hinge, put it in a few key places around the structure, and that lets us scan 02:57.000 --> 03:03.560 around samples automatically. And the early paper was really focused on the mechanical 03:03.560 --> 03:08.440 performance of it, the drift, the repeatability, the reliability and the precision. But 03:08.440 --> 03:14.640 it showed that this is an approach that does work, you can use just the repeatable bending 03:14.640 --> 03:20.360 of plastic to position something accurately enough that you can use it as a microscope. 03:20.360 --> 03:26.280 And from there, the research into open flexure really started. So, we wanted to use the 03:26.320 --> 03:32.400 Raspberry Pi camera for quite a few reasons, but it is, I think, fair to say, and I don't 03:32.400 --> 03:36.080 think anyone here would disagree with me, not exactly what the Pi camera was designed 03:36.080 --> 03:42.440 for. And so, the early question was, well, is this going to work? Can we calibrate the, 03:42.440 --> 03:51.200 it's okay? Can we calibrate the collar to make it a true recreation of the sample when 03:51.200 --> 03:55.080 we're not looking at something that's 2, 5, 10 meters away, we'll look at something that 03:55.080 --> 04:01.400 at times is only a couple of microns away. Here you can see on the left of the before, 04:01.400 --> 04:07.600 I don't know how well the projector is showing this, of applying green, blue and red sample. 04:07.600 --> 04:15.440 This is after calibration. And also, how are we going to do this in a way that's not just 04:15.440 --> 04:22.120 open but repeatable? And, done in a way where we can go through our history, and we learned 04:22.120 --> 04:27.120 a lot of lessons from open software development. Our microscope is designed in open 04:27.120 --> 04:31.640 SD card. It's a programming based CAD package. That means that we get the same version 04:31.640 --> 04:38.280 control as any software development team. It means that when we change the position of 04:38.280 --> 04:44.080 something or the height, we can see that in the history as physical code that's changed 04:44.080 --> 04:51.600 to move the physical parts. That also means that we can do what we called hardops, so using 04:51.600 --> 04:59.080 the GitLab CI to automate a lot of hardware steps. So, mad as this kind of sounds as 04:59.080 --> 05:05.280 a phrase, we link to our hardware. We automatically rebuild it every time something's 05:05.280 --> 05:12.680 pushed or merged. As part of that pipeline as well, we automatically redo all of the vendors 05:12.680 --> 05:17.120 that are in our instructions. It sounds like quite a cute line to say we've spent more 05:17.120 --> 05:21.960 time on our documentation than we have on the physical design, but I can show you the 05:21.960 --> 05:27.760 time sheets, it is 100% true. A big part of that is making sure that we're reaching as many 05:27.760 --> 05:32.720 people as possible who aren't confident engineers or builders. And one thing that we've 05:32.720 --> 05:38.520 got is feedback quite often is the photos in your instructions don't match the parts that 05:38.520 --> 05:43.760 I've got, which was often, you know, we changed the size of a gear or the angle of a hole, 05:43.760 --> 05:49.040 something like that. It's one of the fundamental we didn't matter to the build, but still 05:49.040 --> 05:56.440 people who weren't confident, would he be able, I'll just say quite often. Instead, when 05:56.440 --> 06:01.760 our automatic we rebuild all of our vendors that are in the instructions every time there's 06:01.760 --> 06:09.560 a change, and as part of that, we also have the release on the outside and the GitHash to build 06:09.560 --> 06:15.560 those parts on the inside. That means that we can make sure that everyone's on the same page 06:15.560 --> 06:21.080 with the hardware, it means that if I'm working with a collaborator who swears 100% absolutely 06:21.080 --> 06:28.240 that they did print the right part, I get to go in and check. So it's a popular feature for 06:28.240 --> 06:34.520 me, but some of the collaborators may be not so much. We also worked at how we can compensate 06:34.520 --> 06:41.680 for some of the inherent limits of this approach to building a microscope. The hardware, 06:41.680 --> 06:47.000 I'm sorry to say, is not as precise as a half million pound-side scanner straight out 06:47.000 --> 06:51.120 of the box. Some people come up to our store and tell us that as if it's some sort of 06:51.120 --> 06:59.200 gotcha. But we listen quite smart with this digital microscope, and that's, we put a sense 06:59.200 --> 07:06.400 of in there. So that we can monitor how successful it is moving and working. We're using 07:06.400 --> 07:11.640 the pie cover. So if the things like this, we can, when looking at blood smears, check our 07:11.640 --> 07:16.640 focus as we're going. This is looking at a blood scan on the upper plate of microscope, 07:16.640 --> 07:23.840 when you're 10 microns out of position, 5, 3, 2, 1, and in focus. We know what this looks 07:23.840 --> 07:29.520 like. We can self-calibrate our movements now scanning as we're going rather try to solve 07:29.520 --> 07:42.320 everything at hardware. Working on the software, we align it with the web of things inside 07:42.320 --> 07:47.920 every microscope is a Raspberry Pi that spins up a web server and controlling the hardware 07:47.920 --> 07:52.480 as a simple HTTP request. That means that a lot of people have got in touch saying that 07:52.480 --> 07:56.800 they only wanted part of the microscope and wanted to bring some of them in hardware and 07:56.800 --> 08:01.160 everything into faces seamlessly. It also meant that someone got in touch saying that they 08:01.160 --> 08:06.600 had done open-flight surveillance control for example weaving the microscope in a few 08:06.600 --> 08:13.240 in the cupboard, or an incubator, and blockly. So being in programming to people who aren't 08:13.240 --> 08:21.080 comfortable with programming. We also have some more dead links and two sort of spin-out projects 08:21.080 --> 08:26.600 to the delta stage and the block stage. Same approach in terms of how far can you take 08:26.600 --> 08:32.040 flexure-based 3D printing. All of the same approach of how can we make this as open as 08:32.040 --> 08:38.760 accessible as possible. We published our version 6 of the microscope paper a few years ago 08:38.760 --> 08:45.640 now where we're talking about all the capabilities things like for the SNS imaging, reflection, 08:45.720 --> 08:51.160 transmission, all the different imagery modes and things that we've developed over the previous 08:51.160 --> 09:03.400 few years. This is rough. Well, it means I won't run over time. And then we also shared 09:03.400 --> 09:08.920 a lot of our experiences because we were co-developing this with a team in Tanzania and we do 09:09.000 --> 09:18.840 mean co-developing. And that is in some ways very, very challenging. And some of the lessons 09:18.840 --> 09:25.320 learned about how to co-develop something with partners in Tanzania when, for example, it's March 09:25.320 --> 09:31.880 2020 and your fight has just been cancelled. We were over there just as Heathrow Airport shut 09:31.880 --> 09:38.840 down during COVID so that was about the most stressful work travel I've ever done. And that 09:39.800 --> 09:46.520 we've come to this a 3D printed automated microscope with customizable optics, with different 09:46.520 --> 09:57.800 imaging modes, running a fast API server. And the first rule of talks in engineering is never 09:57.800 --> 10:02.360 given live demo. But considering how well my slides have gone, maybe this is for the best. 10:03.560 --> 10:08.280 So this is a live stream from the microscope. You can either trust me that it's live or I can do this. 10:09.160 --> 10:16.760 And if anyone thinks that I've time to that perfectly, you're more paranoid than me. 10:18.200 --> 10:22.440 But this is the research that's going on to the microscope. Now at the University of Glasgow, 10:22.440 --> 10:27.320 and a few other research places, things like the auto focus that I've just done, clicking around 10:27.320 --> 10:35.160 the sample, capturing images, calibrating everything. So if, for example, you just flown with this 10:35.160 --> 10:40.760 microscope through Edinburgh Airport and security weren't too sure about it. And they decided to test it 10:40.760 --> 10:46.920 by giving it a little shake. We could also calibrate for that. That also means that if you're not confident 10:46.920 --> 10:51.800 in your printer or in the parts that you've got, this self-calibration runs automatically, 10:51.800 --> 10:56.120 giving you a report on whether or not the stage is moving as expected, whether or not the camera 10:56.120 --> 11:01.320 can calibrate properly. One of the big projects that we want to work towards the next couple of 11:01.320 --> 11:06.840 years is a fun calibration report. So when we launch version 7 of the microscope, people could 11:06.840 --> 11:13.480 run this calibration, send us the report and we can start mapping what the limits are. Is it 11:13.480 --> 11:20.600 in certain areas where people use a different supplier or our instructions are clear enough to 11:20.600 --> 11:26.040 people with English as a second language, things like that, mapping up what are the actual challenges 11:26.040 --> 11:31.000 around reproducing hardware around the world? Now it's hope that that calibration works. 11:39.160 --> 11:44.440 There's stuff. From here we can also do automated scanning and this has been a big focus of our 11:44.440 --> 11:51.320 work, the last few last couple of years, not just scanning a single field of view, but doing an 11:51.320 --> 11:56.920 automated scan where we can capture a VFocus and move around building up huge stitched images, 11:58.280 --> 12:04.360 like this, for you automatically getting the kind of detail you need to see inside a cell, 12:04.360 --> 12:12.120 it looks even better not on a projector. Variable for hobbyist, valuable for engineers, 12:12.120 --> 12:16.120 potentially invaluable for people like pathologists who are overworked and don't have the 12:16.120 --> 12:20.920 funding to hire the number of people they need to be scanning samples and spending time with patients. 12:22.120 --> 12:28.040 And that kind of would be it, as I'm having the five minute sign, where you've got me, 12:29.320 --> 12:37.720 but there's also all the people we've got in touch with broken lengths as well. Not one 12:37.720 --> 12:45.000 after we launch this, something started happening because we did something that seems insane to 12:45.080 --> 12:51.880 anyone, maybe not in this room, which is that we made this open source and as accessible as we 12:51.880 --> 12:56.120 could as well documented as we could and it wasn't long until a research group at a different 12:56.120 --> 13:01.320 university contacted us saying, hey, we've found your design, we've modified it to do super 13:01.320 --> 13:10.680 resolution microscopy. Cool. Or a team out in Japan got in touch saying, hi, we've modified your 13:10.680 --> 13:15.880 design for structure the illumination. So physically moving or grading around a filter out of 13:15.880 --> 13:20.520 adding out of focus light to get resolution higher than you would be able to on this otherwise. 13:21.320 --> 13:28.200 We've done that, we've taken all the data, we've published the paper and we thought we'd get in touch, 13:28.200 --> 13:37.160 we hope you don't mind. We didn't. For how they were able to modify the hardware so easily, 13:37.160 --> 13:45.160 I should put you towards Julian's talk this afternoon at 140. People also got in touch sending us pictures 13:45.160 --> 13:50.440 of, come by the store later, I'll show you the actual pictures. People got in touch showing the 13:50.440 --> 13:56.360 workshops, they've run with the education workshops, engineering workshops, also commerce workshops 13:56.360 --> 14:01.480 setting up the idea that people can sell this and turn a profit and support the community that needs 14:01.560 --> 14:08.680 more hardware, more technicians, more people using microscopes. And then people also started getting 14:08.680 --> 14:13.800 in touch saying that they'd used our microscope, not because they were interested in a 3D printed 14:13.800 --> 14:19.720 microscope, but because they needed a microscope and we were the best option available to them for 14:19.720 --> 14:25.800 one reason or another. And that became a bit of a transform for this where we've started to feel 14:25.800 --> 14:30.360 like we were just part of the scientific landscape, rather than being the novelty for someone who 14:30.360 --> 14:38.600 wanted to try out 3D printing, we were part of the microscopy world. People also got in touch 14:38.600 --> 14:43.000 a lot about healthcare applications, so a team in the Philippines got in touch saying that we've 14:43.000 --> 14:49.160 shown that your device has the ability to look at samples for tuberculosis, heart disease, 14:49.160 --> 14:55.400 and multiple other conditions, and it works. Dr Daniel Rosen, a professor of pathology over 14:55.400 --> 15:00.120 in the United States, is one of the big people who's been driving this project forwards, 15:00.120 --> 15:04.680 no engineering background at all, but I think at this point now is 6 3D printers in his shed. 15:07.080 --> 15:12.360 And he's been telling us, you know, you guys don't know what you made, this wasn't just a microscope, 15:12.360 --> 15:17.400 this is a slide scanner, digital slide scanner, which is prohibitively expensive for 15:17.400 --> 15:24.680 the majority of the world. He worked with us, he's published papers using our microscope to image 15:24.680 --> 15:29.560 cancer in the esophagus, and was also the driving force between a recent trip we had to 15:29.560 --> 15:36.600 so power. Hospital DMR in so power is the main cancer center for the country, and as part of that 15:36.600 --> 15:42.680 work, they buy trucks, strip out the back of them, and turn them into a lab, and that lab, 15:42.680 --> 15:49.080 goes and parks up wherever it's needed across the country, and where I say goes wherever it's needed, 15:50.120 --> 15:54.040 there is an incredible picture there, which has not worked with them floating one of these on a 15:54.120 --> 15:59.560 barge down the Amazon. So if anyone ever asks this open flex, if available on Amazon, 16:03.560 --> 16:08.920 we also use this platform for open pathology education, because the digital images that come 16:08.920 --> 16:14.200 off this microscope can be used for education, and it means that some of our collaborators in 16:14.200 --> 16:20.200 underserved regions can now run online lectures on a digital scan of the samples from cases 16:20.200 --> 16:27.240 that they see every day in their practice. The alternative is everybody gets trained on the same 16:27.240 --> 16:32.200 digital scans from areas of Europe and North America. It doesn't reflect the sample, 16:32.200 --> 16:36.680 doesn't reflect the conditions, doesn't reflect the sample preparation, that the majority of 16:36.680 --> 16:42.840 doctors around the world are looking into. We set up a forum so that people can start sharing their 16:42.840 --> 16:51.400 questions and ideas and work with us, and also set up a YouTube search alert, and these are people 16:51.400 --> 16:57.560 that are off and outside of academia, who there's no conventional academic way to ever reach out 16:57.560 --> 17:04.440 to people who had a 3D printer, were happy ordering parts online, and then you're doing things 17:04.440 --> 17:11.400 like machine learning on the pie to label various features of their samples from the forum as well, 17:12.280 --> 17:17.800 literally someone got in touch, I don't know his real name, saying that he was going to adapt 17:17.800 --> 17:23.720 this to do a top atomic force microscopy, to give you an order of scale that's going from looking 17:23.720 --> 17:29.160 at individual cells, to looking at individual atoms on this thing, and we thought, it's a 17:29.160 --> 17:36.600 lot of way idea, and then the schematic started coming out of his apartment in New York, 17:37.160 --> 17:40.840 some of these schematics had to be read on, because his children found them and thought that they'd 17:40.920 --> 17:47.560 help by colouring them in, and he now has a prototype, and just a reminder of his images continue 17:47.560 --> 17:55.720 to fail to load, that's where all of this started, and now we're here and with dozens of other projects 17:55.720 --> 18:00.760 I could tell you about, this is our world map of every country where someone's got an 18:00.760 --> 18:06.360 in touch with us saying we've built or used your microscope, and as much as I wish that this was 18:06.360 --> 18:11.240 me flying to 70 different countries with this in my hand luggage, that's not the case that's 18:11.240 --> 18:16.680 people finding this organically and getting in touch, the map tells half the story, but it's the 18:16.680 --> 18:21.480 people involved, the fact that if you put this out there people will find it and use it for things 18:21.480 --> 18:28.920 that you never thought were possible, or maybe even a good idea, and that's as far as we thought 18:28.920 --> 18:33.560 it would go with that world map with all the different countries coming in and hopefully every 18:33.560 --> 18:40.920 few months will be able to add one or two more, and then on LinkedIn we saw this a couple of months ago 18:42.120 --> 18:48.040 a group looking at immune cells in microgravity needed a microscope that they could adapt, 18:48.040 --> 18:54.040 it had to be lightweight, and it had to be something that was open source so that they could 18:54.040 --> 18:59.880 interface with other components, and so our world map, as of March 2027, is going to be a bit 18:59.880 --> 19:07.320 obsolete, and we're going to be doing a world gift instead. I'll finish with the conclusions from 19:07.320 --> 19:14.280 that first 2016 paper, written before I was on the project, and highlight this microscope is 19:14.280 --> 19:19.160 a useful tool on its own right, and there is much potential we plan to further investigate in the 19:19.160 --> 19:26.840 future, so in ten years essentially nothing has changed. I'd like to thank you all for listening, 19:26.840 --> 19:32.280 I know that I've got a swap that many of you wish that you didn't have to get up for, 19:32.280 --> 19:38.120 we have a stool just along here, and two more talks from members of the team on the educational 19:38.120 --> 19:44.600 applications and on the hardware journey itself. I'm very excited to take any questions, and you can 19:44.600 --> 19:47.000 find it out there until the end of the day. Thank you. 19:47.000 --> 20:07.800 Okay, well, we'll do, okay, thank you. 20:07.800 --> 20:20.680 In Antarctica, is using my requires strong words? The question was, do we know who was 20:20.680 --> 20:29.000 using the microscope in Antarctica? So the users per pixels in Antarctica is pretty good. It was 20:29.000 --> 20:36.120 a biologist a few years ago who wanted to take samples of the ice, triage them out in the field, 20:36.120 --> 20:42.600 so to speak, and then take the samples that looked interesting back to her lab. She's not on the project 20:42.600 --> 20:50.040 any more, but we're always going to be grateful for the work that she did. Is that else? 20:50.040 --> 20:54.440 Yeah, of these companies that sell the microscope. Yes. 20:54.440 --> 20:57.880 Now, Lord, you're skimmed, it's down with everything, and then you sell it to yourselves. 20:57.880 --> 21:00.440 Do you have any arrangements with them in financial arrangements? 21:01.320 --> 21:10.920 We have a, yes, so we do have, as you say, I think it's 12 vendors around the world at this point, 21:10.920 --> 21:17.960 who have got in touch saying that they'd be interested in selling kits for the microscope. 21:20.200 --> 21:23.800 Our license is very clear that we can't stop them, but we also don't want to. 21:24.520 --> 21:31.800 And we have regular meetings with them, we have a memorandum of understanding. Some of them voluntarily 21:31.800 --> 21:39.480 gives some of the profits back to the charity that supports disk growing network, others instead 21:39.480 --> 21:44.520 contributes to the code base, things like that. And then on our website where there's a list of vendors, 21:45.160 --> 21:48.920 we try to give it an idea of what contributions each of them voluntarily make. 21:49.880 --> 21:57.080 But to be honest, the contribution of them selling it is already enough of a contribution, 21:57.080 --> 22:01.880 because I don't like counting out screws and small bags. So, thanks, guys. 22:03.880 --> 22:05.160 Please. 22:05.160 --> 22:10.360 Is there any concession with the other, if you like, of course, a key platform? 22:11.080 --> 22:12.840 Are there any concession for the... 22:17.800 --> 22:25.720 Okay, so a question, first thing in the morning about our API have on and working with other 22:25.720 --> 22:26.840 open microscopy. 22:30.840 --> 22:35.080 I'm not aware of anything with four men go, but what you're doing? 22:36.920 --> 22:37.560 Okay. 22:41.320 --> 22:49.080 It means a bit, but for example, you see two of anyone's familiar with Benedict Deedrick's project. 22:49.080 --> 22:53.800 It's a series of modular cubes that all slot together so you can build an optical path. 22:53.800 --> 22:57.480 As long as you want, we don't quite have the space for that in here. 22:57.480 --> 23:02.360 So a few years ago, we worked with Ben A to print a little adapter plate, 23:02.360 --> 23:06.600 by essentially he bought the optics and we bought the stage, and they interface perfectly. 23:08.200 --> 23:08.760 Anything else? 23:10.360 --> 23:15.960 I'm really inspired to use a new process to another organization with a particular adapter 23:15.960 --> 23:21.560 as create a model for you. Is there a code review for the hardware you built it, how does that work? 23:21.560 --> 23:25.480 Oh, yes, so this code review for the hardware is much as this for the software. 23:27.080 --> 23:31.720 People have folks this and then modify it in different ways, quite often it's things like they only 23:31.720 --> 23:36.920 wanted, perhaps the Z control rather than the XYZ control. Sometimes they'll live as folks, 23:36.920 --> 23:40.280 but other people do get in touch saying that I think this is something that everyone would 23:40.280 --> 23:48.520 be interested in. Our documentation also has a list of customizations and other versions, 23:48.520 --> 23:53.400 which is where some of those live. So our microscope's unusual in that we have the camera in the 23:53.400 --> 23:59.400 base and the illumination at the top. There's a customization for this, which isn't as vigorously 23:59.400 --> 24:07.160 tested where that's inverted, but yeah, the code review for the hardware is a huge part of what we do. 24:07.560 --> 24:09.960 How was the time? 24:09.960 --> 24:13.160 Yeah, thank you so much, thank you so much.