xRAPID is live

After two years of development, xRAPID, my latest app, is finally live — you can find it here. That I know, it’s the first medical diagnostic of a major disease running on an iPhone (and in the App Store): it detects the presence of malaria parasites in red blood cells. There is a lot to say about it, in terms of development and creating a company for it, getting it to work on a portable microscope (you got to purchase it from us), and making the whole package robust. The latter took one full year, but I believe it was worth the effort. I have a full paragraph of what it is after, a few short remarks about pricing and the business model, but I thought I would first explain a bit the genesis here.

Genesis

The idea of detecting malaria on an iPhone came about 2 years ago. I have been in Indonesia for 3.5 years now, and given my current job and the lack of technical resources, was looking to expand my fields of expertise. I happened to see a microphotography of malaria infected blood, and I was toying at that time with some non-linear photo filters. I tried a home-built one, with a few modifications, and was able to segment properly the image to show only the stained parasites. Three hours of work, literally. I then purchased a toy microscope, and tried on a real slide. Same sort of success, with a few caveats. And then it hit me. Complete change of strategy. I have been coding for the best part of the past 25 years, had been toying with iOS development and had a few apps under my belt, I had even worked as a 3rd party dev for a while.

The fundamental change, for me, was answering this question: ”I have been amazed and have verified the claim that I’m carrying a computer in my pocket since 2007, why haven’t I done anything with it?”. And by anything, I meant not YA(F)SN ( Yet Another (Censored) Social Network). Or a game, or productivity app — I’ve done all this, it’s not that complicated honestly. No, the real question was: what about the hard stuff, the stuff I generally do only on a workstation? Hard engineering stuff? Off I went, and realized quite quickly that what we said was true. It turned out that fairly intensive computing could be done on the iPhone, particularly related to image analysis, very much so when you could stick to GPU image processing.

You know the process if you are an engineer: get to work on something, split the problem into tiny chunks that you think can be resolved, and start working them one at a time. Each time thinking to yourself, I cannot believe that this is working. At the same time I was progressing, the kit was progressing. I got it to sort of work for a proof-of-concept on an iPhone 4S, skipped the iPhone 5 as I was in the bulk of the filter development, and then got to the point where I was sure it would work on the iPhone 5S. It was a quantum leap in terms of performance. It also really helped that some routines (BLAS etc) became available during the development.

xRAPID — a protable microscope to which an iPhone is hooked, the app does the rest of the magic…

From the early prototype to the product: the company!

I showed this to a friend, who talked to one of his friends, who contacted me. I had been in touch with Jean since the school-in-a-box project, he was the one who sold me the kit. There was not enough traction on this product so he had decided to pause it, and being a natural entrepreneur, was looking for his next jig. This is when the magic appeared. We had two 2-hours long chats, he did his homework, and decided that there was a market for malaria diagnostics on iPhone. We incorporated the company in London in a day ( take that, France. Seriously, any country that does not enable the creation of a company in an hour is doomed), and decided to put all other activities on the side while going for it. Looking back, we have made remarkable achievements in twelve months:

  • create a company
  • create a prototype and demonstration
  • secure a round a seed funding
  • test 4 different microscopy solutions
  • source and adapt to our needs a portable microscope
  • get the app from a proof of concept with barely workable UI to a robust field-tested solution
  • qualify the app on lab slides
  • establish partnerships with two leading malaria centers (London School of Hygiene and Tropical Medicine, and Paris’ La Pitié-Salpêtrière)
  • get a qualification as a parasitologist
  • qualify the app and the kit with >1000 hours of field test
  • qualify the method with over 250 test slides
  • establish the statistical independence of results from the kit, the app and the operator
  • write manuals and establish all the material that needs to be delivered with the kit, including training
  • established a business model and several offers, started sales and established a sales team
  • hired our first employees
  • oh yeah, get the app through the App Review process, in a category I had never tried before (Medical).

Now we’re not done by far, there is much to go on, such as FDA and regulations approval, but: it’s working and it’s here, available. I believe we can be proud.

xRAPID: A Mobile Automated Malaria Field Diagnostic

So, we have released the first fully automated diagnostic method and apparatus for rapid malaria detection using a microscope attachment to a mobile phone. While various attempts have been made at computerized image analysis for detecting malaria in thin and thick smears, xRAPID achieves in-phone, off-network diagnostic capability. The main attributes are:

  • a miniaturized back-lit upright microscope, with x, y and z independent displacement
  • an attachment and support for a mobile phone (iPhone 5S, 6, 6+ compatible)
  • an app for automated diagnostic of the four main types of plasmodium encountered in the field: falciparum, vivax, ovalae and malariae (including co-infection)
  • a local patient database with networking capabilities, including strong encryption and/or anonymized data
  • geotagging of data with live/asynchronous reporting to a central server.

The microscope achieves maximum portability, with dimensions of 13x11x5cm3 and weight of 480g, while retaining all the capabilities of a standard upright microscope, with 100, 400 and 600x magnifications. The diagnostic of Giemsa-stained specimen is based on

  • a combination of non-linear chromatic filters,
  • a size analysis,
  • a measurement of color and granularity of cells,
  • a digital pattern recognition and artificial intelligence algorithm.

The analysis does not require a network connection. A single field of view is analyzed in a second, a quantitative test with a parasite count accurate to 0.1% in about a minute ( conservative value by all means, we are working hard to make it down to 0.01%. And we are not far from this figure).

A note on the business

I love solving problems, even more so when they have the potential to change or save lives. Now, there is a huge difference between a simple academic exercise — solve the problem, demonstrate that the solution works with a few slides, write a paper and be done with it — and making it really useful. We should define useful at some point, but picture this: in a world (a third world, admittedly) connected to communication means, i.e. the Internet, with a good spatial spread of xRapid we can use the iPhone to report live anonymized diagnostic results. It is not a feature of the far future, we do it right now. Cache a simple short string of characters, encrypted, and beam it back to a central server whenever there is enough connection to send an SMS. We believe that xRapid is a game changer right now, just imagine the near future with me: geo-localized live data on infections. It basically means that we will have displaced the problem from the medical to the sanitation domain. Of course, once we know where an infection is localized and how it is spreading, we can act upon it within 48 hours. Couple that with weather data, and you can dispatch a sanitary force, equipped with the right insecticides, and get the region free of the problem: being right and quick is essential, as you are trying to break the reproduction cycle of the parasite.

There is no way to achieving this vision without having a proper business to support it. Even if it were possible to give it all away, open-source all the tech and wish for the best, I can assure you nothing would happen out of it. There are many reasons for this, but the main remains: nobody cares. A first world solution to solve a third world problem? Come on. Nobody will believe it can work until we have done the maths for them and demonstrated that not only does it work, but also it is faster, cheaper and more reliable.

So, we made it into a business. And tried to be fair while doing so. It means putting a fair price on the technology. Fair is pricing it as low as we can, while making enough money that the operation is sustainable from day one, and can be expanded. You see, malaria is not the only disease we can do. I have been exploring others, actually. A large body of bacteriology can be dealt with in the same way, for instance. The full blood count can be given (hint: we are not that far from getting it within the highest confidence interval, it will still require a lot of testing though). I am fairly confident that the Tse-Tse protozoa can be identified and diagnosed properly as well. We have also experienced with the Loa Loa worm. And so on and so forth. A lot of the traditional biology can be translated into automated systems leveraging the analytical capabilities of a mobile phone, and my personal feeling is that the basic technology we now have can be used as a platform, while the process we have followed can be generalized to expand to other diseases.

To cut discussions short: we are pricing the individual diagnostics below a dollar. Depending upon the customer needs, the kit can be either acquired or rented, just like a mobile phone… meaning you can get it on contract or as a pay-as-you-go. The former purchase large numbers of tests and get the kit for free, while the latter purchase a minimal number of tests but purchase the kit. In the long term and for a large number of tests the contract is more advantageous, of course, and we tried not to replicate the same mistakes as telecom carriers (which I truly hate) and be as flexible as can be — contracts can be upgraded on the fly, etc… The one constant: under a dollar (we should be roughly costing a half to a tenth of existing rapid diagnostic test kits).

Conclusion

We are of course happy to have the app and kit in the wild — pretty much so as we have shipped to Papua, which is as wild as it gets. And I am particularly happy that we have a bette offer than the traditional RDTs (rapid diagnostic tests): we are faster, cheaper, specific (xRAPID determines the type of parasite and parasite count), and do not degrade over time, temperature and humidity.

One of the most exciting parts of the job is definitely that the kit keeps getting better, we have a large body of references now that contribute to making testing much faster and efficient, and therefore enable faster turn-out for development. The coming weeks are going to be very interesting as we receive feedback about the app being used in the wild, and quickly iterate to increase the performance of the app. Exciting times, really.