,
24 tweets,
12 min read
Read on Twitter
(1/n) Latest pre-print from lab! Octopi; open configurable high-throughput imaging platform. Led by an incredible graduate student @hongquan_li - we crack puzzle of automated low-cost malaria diagnosis. Remember, malaria still kills ~1/2million people/year biorxiv.org/content/10.110…
(2/n)The empathy & wicked building skills that @hongquan_li brings to this project is innumerable. He has been the most amazing partner. Many people helped including Hazel (malaria samples), Maxime and Lucas (AI), Thibaut who traveled with me. Here is Hongquan supervising a build 
(3/n) Let me first tell you a story of why we built Octopi. Imagine you are lab tech in a remote hospital with hundreds of patients waiting all day outside the clinic everyday. They all want to know - are they sick from a deadly disease like malaria. You work 10hrs/day non-stop.
(4/n).. barely eating, trying to get through samples. Off course you don't have electricity. Not enough hours in the day. And the patients keep coming. This is the life of incredible staff and lab technicians like Praveen and Durga I met in Kalahandi, Orissa. @Swasthya_Swaraj
(5/n) This is not isolated story, but reality of where malaria hides, across the world. 3.4 billion people in 92 countries are at risk, ~200 million cases/year. RDTs have lots of problems(see paper). Even with incredible individuals fighting diseases, we don't have the tools..
(5.5/n) Oh and did I say that number of malaria cases are rising in the last couple of years. That can not be good!
(6/n) In 2015 trip to Uganda, I realized that we need low cost but high-throughput and flexible tools to fight these deadly disease. Tools need to adopt to the context (large hospital or small village) - like "transformers". Listening to endless accounts of lab techs in field..
(7/n) ..Octopi was born. At a price point of ~$250 in parts, we demonstrate automated quantification of malaria using multi-modal channels - fluorescence, bright field and spectral imaging. Operate on phone charger, Octopi scans 1.5 to 5 million cells/minute - picking parasites.
(8/n) In late 2015, my first inspiration came from high throughput scanning in DVD drives. We got hundreds of DVD drives in the lab, to better understand scanning, autofocus mechanisms that are mass produced for nothing - and now I have the largest graveyard of drive optics.
(9/n) @hongquan_li (infinitely talented) joined the project, & his rotation project was to reverse engineer DVD player. We explored 100's of scanning mechanisms. it helps that @hongquan_li knows even the most obscure parts catalogs. In the end, we decided to build from scratch..
(10/n) Key idea in paper is completely modular, stackable design - every module magnetically couples in & out. So if you want to diagnose malaria, switch to malaria head. Want to detect Tb - just swap the head. And anyone can build new modules to expand capability since its open.
(11/n) Next we built a scanning & autofocus module. It took a while, but now we can make scanning stage with flatness of almost 300nm for ~$65 in parts. we built different autofocus modules; here is one in action. Flatness matters a lot to reduce imaging time in clinical context.
(12/n) Octopi can scan between 1.5 to 5 million cells/minute, both in bright field and fluorescence to make giga pixel images. It can adapt to take both low/high mag objectives. So a simple finger prick blood thin smeared, dipped in a stain and you are ready to image.
(13/n) Our next breakthrough came when @hongquan_li first saw a 10nm spectral shift for nucleic acid stained parasites compared to platelets. This was a big moment for both of us. Lot more to come on this front, but its likely DNA/RNA ratio in different cells varies.
(14/n) This is a big deal since using RGGB Bayer filters on color sensors &low mag, low-cost air (10x) objective, we can detect parasites. In only 16 FOV, with bright-field, fluorescence and spectral information combined - we can do 90% specificity/sensitivity for 50 parasites/ul
(15/n) And since Octopi is modular, we can also image the same parasites at high mag - but we don't have to keep searching for them. In fact, we built one module that does both low and high mag in tandem! Here are some nice ring-stage Plasmodium falciparum. Thx Ellen Yeh lab!
(16/n) Off course, Octopi applies to many other applications in digital pathology, Tb.. but also for basic science. I often call Octopi - big brother of #Foldscope. With automated imaging, tracking, all built in rugged portable unit, we have also been doing environmental imaging.
(17/n) We will be posting all the CAD files, assembly instructions, code, field data online - we are just cleaning git hub repositories. Keep an eye on this space for lot more to come. Also; the final instrument is machined - although we made 3D printed versions too.
(18/n)If you have been reading so far, I have an ask. If you like robotics, biology & want to tackle these problems; join forces. Build DIY tools, apply them to problems you care about. Make new Octopi modules, teach someone, start Octopi club.. microscopic world is for everyone.
(19/n) I imagine a future where a network of low-cost automated “smart” microscopes capable of identifying pathogens arm community health workers - and work hand in hand. Its possible through collective efforts from community of research, developers, clinicians and philanthropy.
(20/n) We are headed off to the field again shortly to do further testing. Post that trip, we will be announcing a new program (remember foldscope program) to scale up deployment & requesting field collaborator, & community engagement. Starting with 100 Octopi deployment first..
(21/n) Thx to @gatesfoundation @HHMINEWS @NSF_BIO that supported financially. But we have ong way to go, specially scaling up these tools to match a dream of robotic network of microscopes. If you want to support scale-up, field testing & 100 Octopi deployment, we need your help.
(22/n) I also want to thank so many community health workers and NGO and government workers that fight infectious diseases. The wisdom and experience in these communities is untapped. We learned a lot by spending time with @Swasthya_Swaraj @pivotmadagascar and so many more..
(23/n) On lighter note (after my longest twitter rant); you might wonder why we call it "Octopi". Octopus are smart, a network of them can take down ships (evidence below), are versatile/flexible, & have freaking eight arms (modular). Also, my kidos love Octopus. Suggest a logo!
