Over the next 4 episodes, we’re going to take you on a tour of the CZ Biohub in San Francisco where scientists are working to cure, prevent, or manage disease by the end of the century.

Every year, 400,000 people die of malaria – usually children in impoverished areas. In the first episode of our Biohub series, you’ll meet Paul Lebel, a member of the CZ Biohub Bioengineering team, who is helping to create a device that uses machine learning to accelerate and improve the process for identifying malaria-infected red blood cells. If the project is successful, the hope is that they can diagnose the disease faster and help save lives.

Learn more about the Bioengineering team at CZ Biohub:
https://www.czbiohub.org/bioengineering/

Transcript:

0:00:00 – (Nate): For the next couple of episodes, we’re going to be exploring a place in San Francisco, California, known as the Chan Zuckerberg Biohub.

0:00:08 – (Sandy Schmid): The Chan Zuckerberg Biohub is a small private medical research institute that’s really founded to try to understand human biology, understand the mechanisms of disease, and do that in partnership with amazing nearby universities, Stanford, UCSF, and Berkeley, so that we can leverage our cumulative and complementary knowledge and technology to advance science as quickly as possible.

0:00:37 – (Nate): This is Sandy Schmid.

0:00:38 – (Sandy Schmid): I’m the Chief Scientific Officer here at the Chan Zugerberg Biohub. And it’s great to have you visit.

0:00:44 – (Nate): At the Biohub, scientists are working hard and working fast to cure, prevent, or manage all diseases by the end of the century.

0:00:53 – (Sandy Schmid): Absolutely right. And in a different way, in a way that creates a community so that all scientists are rowing in the same direction and working together towards accomplishing those goals, communicating their information and advances really rapidly and openly and discussing and collaborating. So it’s getting to those results in a new way that’s going to be much more impactful and effective than has previously been on.

0:01:19 – (Nate): Well, we’ve got a great tour lined up, and our first stop is the bioengineering lab.

0:01:31 – (Nate): Hello, everyone, and welcome to another episode of The Show About Science. This is your host, Nate. We’re about to enter the lab, and you’re probably wondering, hey, Nate, what even is bioengineering?

0:01:46 – (Nate): Yeah, so what even is bioengineering?

0:01:50 – (Paul Lebel): Yeah, that’s a really great question. Bioengineering can mean a lot of different things to a lot of people. It can either mean you’re actually engineering new biological systems. That’s not what our team does, actually. But what we do, in contrast, is applying mechanical engineering, electrical engineering, optics design, fluidics design, software, all of these different fields. We leverage them to build tools for biology.

0:02:15 – (Nate): This is Paul Label, a bioengineer at the CZ Biohub, and he’s using those tools to fight malaria. And the first question that I thought of was, well, why malaria?

0:02:28 – (Paul Lebel): That’s a fantastic question. I would almost counterback and say, why aren’t more people focusing on malaria? And the reason I would ask that is because every year, 400,000 people die of malaria. And these are usually children and usually in impoverished areas. So if you look at COVID deaths and how much effort we spent on something like COVID, there was a huge tragedy and loss of life. But this happens year over year with malaria.

0:02:52 – (Paul Lebel): And if we can help diagnose it and get people treated faster, we would be saving lives. So I don’t have a better answer.

0:02:59 – (Nate): Than that, but that’s a pretty good answer.

0:03:02 – (Paul Lebel): There is a lot of useful things to work on. I think this is one of them.

0:03:07 – (Nate): Malaria is a parasite spread to people by mosquito bites. Now, it’s pretty easy to see under a microscope. Here’s how they do it.

0:03:16 – (Paul Lebel): So the way this has normally been done and been done for literally 100 years is people take a drop of blood from someone’s fingertip, they smear it on a glass slide, blood’s all smeared out, and then they have to use a bunch of chemicals and reagents to process that blood slide. That whole thing takes time, it takes skill, it takes training, all this stuff.

0:03:38 – (Nate): And then once the slides are ready, somebody actually has to go in and find the malaria.

0:03:44 – (Paul Lebel): Yeah. Then what they do is they take that slide and they put it on a regular microscope and they look at it with their eyes and they literally count how many malaria parasites they see in the person’s blood.

0:03:56 – (Nate): Day after day, these malaria spotters are constantly pouring through these slides looking for malaria parasites among otherwise healthy red blood cells.

0:04:06 – (Paul Lebel): So that, you can imagine, is error prone and very tiring. Some of these technicians run 20 to 30 slides per day, so they’re spending 12 [to] 14 hours just literally doing that. So what we’re trying to do is automate that and sort of cut out a bunch of the steps in between to make it faster and more reliable.

0:04:24 – (Nate): Sitting on the desk in front of me is a prototype for that faster, more reliable microscope. It’s a 3D printed black box, roughly the size of a pineapple, and it doesn’t look anything like a normal microscope, nor does it use slides.

0:04:41 – (Paul Lebel): So we basically take the blood almost straight from the fingertip and it looks at the blood as it flows it past the microscope and uses machine learning to look at blood cells and decide if there’s parasites in them. So basically removing all the error prone steps of the way it’s done now.

0:05:02 – (Nate): Yeah. So how much time does this take versus what they’re currently doing?

0:05:10 – (Paul Lebel): Yeah, so currently a technician, it would take them between half an hour and 45 minutes to prepare the samples to look at on the microscope, and then it would take them depending on the sample, it could take them a very short amount of time. If it’s heavily, heavily parasitized, it’s quick. But if the person’s actually healthy, it takes a lot longer because they don’t see as many parasites or they see no parasites. So then that would take them up to 20 minutes or half an hour just looking down the microscope to say, oh, this person’s actually healthy, whereas this takes 20 minutes end to end.

0:05:40 – (Paul Lebel): So you collect the sample and then 20 minutes later, you would have an answer.

0:05:45 – (Nate): So what are you using to make this and make it work?

0:05:51 – (Paul Lebel): Yeah, so we’re using a lot of different technologies that are now available and help us manufacture things and prototype them in house here quickly and efficiently. So we have a lot of 3D printed parts on this microscope. All these black parts are 3D printed. We have a low cost computer inside of here called the Raspberry Pi. We have a really neat device called an Intel Neural Compute Stick. So that allows us to do deep learning cheap and fast, instead of using a big computer server in some centralized building.

0:06:23 – (Paul Lebel): We actually have everything on board here that we need. We have a microscope. Inside of here, there’s a camera. So we’ve done a lot of innovation in house to figure out how to build this kind of thing really cheap and make it work.

0:06:38 – (Nate): Now that I understood how it worked, I was ready to see it in action. But there was one microscopic complication.

0:06:46 – (Paul Lebel): So we don’t have malaria on this microscope currently, that wouldn’t be safe in this lab. But we have a video that we acquired on the same microscope that we loaded on here. So these dark things, those are red blood cells. And this one, you see how there’s two objects in there that shouldn’t be there? Those are actually malaria parasites. Those are younger ones. And this one is a really large, mature one.

0:07:10 – (Nate): And that one doesn’t really look like a red blood cell. Is that like a virus or a parasite or something?

0:07:18 – (Paul Lebel): That’s a good observation. It is a red blood cell, actually. But that doesn’t look like one. I agree. It’s called an aquinocyte. So in your body, you’re always making new blood cells, and there’s always blood cells that get old and get recycled. When you take blood that’s been out of someone’s body for a while, it starts to age, and the aging cells actually start to form these spikes on them, and they turn into, like, these spiky balls looking object called Aquinocytes.

0:07:45 – (Nate): Yes.

0:07:46 – (Paul Lebel): Good question.

0:07:47 – (Nate): Yeah, that one looks very different.

0:07:49 – (Paul Lebel): Absolutely.

0:07:51 – (Nate): On the screen, I could see frame after frame of red blood cells going by.

0:07:56 – (Paul Lebel): This is the way it would look if you’re actively acquiring data, as if you had a real sample on here. So the camera would be streaming images. We’ve actually slowed this down. It goes really, really fast in practice because we acquire so much.

0:08:08 – (Nate): Still going very fast, I’d say.

0:08:11 – (Paul Lebel): Yeah. This is probably tenfold slower than it would be going in practice.

0:08:16 – (Nate): So there are a bunch of red blood cells, and I can make out a couple of Aquinocytes too, that are just, like, going past the screen, I’d say fairly quickly, but I guess not as quickly for the computer. But unfortunately, I am not a computer.

0:08:41 – (Paul Lebel): Same here.

0:08:42 – (Nate): Yeah. So what would it be looking for in this? And how would it know whether this stream of data is healthy or unhealthy?

0:08:54 – (Paul Lebel): It’s a really good question. So we have object detection networks. These are machine learning models that can look at each image frame and decide where it finds objects and what types of objects those are. And so there’s actually many ways of doing that. But basically, it locates regions where it thinks there’s objects, and then it applies a couple of different classification layers to decide what those objects actually are.

0:09:20 – (Paul Lebel): And then we can tally those results. After acquiring 20, 30,000 raw images, then you can tally those results and say how many of these cells actually are sick and how many of them are healthy.

0:09:31 – (Nate): Yeah. Okay. I think I just saw, like, a couple of infected looking cells right there.

0:09:38 – (Paul Lebel): Yeah, you’re right. This one right here, that is one. You want to help with our annotation efforts? See, we can recruit you.

0:09:47 – (Nate): I feel like I’m going pretty slowly, so that might not happen anytime soon. But.

0:10:00 – (Nate): After seeing it in the lab, I was pretty certain it was ready to use, but I wanted to ask just to be sure.

0:10:08 – (Nate): So is this, like, fit to use.

0:10:10 – (Paul Lebel): Currently or not at all? We’re in an R and D sort of phase. We are currently preparing sort of a mad dash sprint to finish the whole system to collect data in Uganda. So we’re at the stage we’re going to deploy the microscopes and basically do a big data collection study. So we’re collecting the century old Giemsa blood smear method. So they’re going to do the manual microscope method, but they’re also going to use a method called PCR, and that’s a highly, highly sensitive tool.

0:10:42 – (Paul Lebel): It’s just not cost effective for routine diagnostics. So in a case like this, we’re funding a study, we’re doing a research study. We can do PCR at the same time and get a really accurate answer for any particular patient to compare our system with.

0:11:02 – (Nate): So do you have any advice for kids who might be interested in working in a similar field?

0:11:10 – (Paul Lebel): It’s been a long time since I’ve been a kid now, but be curious. Obviously, it’s never too early to start being interested in pursuing things. My advice is that you don’t have to find the perfect thing, but there’s just so much interesting and also so many impactful things you can do that there’s just tons of opportunities.

0:11:29 – (Nate): All right.

0:11:30 – (Paul Lebel): Yeah.

0:11:30 – (Nate): Thank you.

0:11:31 – (Paul Lebel): Yeah. I hope you enjoy the tour and the rest of your visit, and feel free to come back and ask more questions if you want.

0:11:39 – (Nate): All right, on the next episode, we will be back at the CZ Biohub, and we will be asking a very important question. How is the cell constructed? Our guest, Manu Leonetti, is going to be walking us through how they’re trying to map the cell using genetic glow sticks. Make sure you’re subscribed because you’re not going to want to miss the next episode. There you have it, folks. The Show About Science is complete.

0:12:08 – (Nate): Paul wants to let you all know that science is a team sport and that he’s just one member of an amazing team. Thanks to everyone at the lab for letting us record there. This episode wouldn’t have been possible without everyone at CZI and the CZ Biohub. Extra special thanks to Patricia Condon, Pete Farley, Jeff McGregor, Dale Ramos, and Sandy Schmid. You are the best, and thank you for the snacks. Okay, dad, you can shut the recording off.