Artificial intelligence and smartphones are being trialled to stamp out malaria in remote parts of the Philippines.
Technology originally developed for marketers to find Nike ticks and Adidas three-stripes on Instagram has been repurposed to detect the disease in blood samples. A person with no medical training could take a picture of a blood sample with their smartphone, and immediately know whether malaria is present or not.
‘If this works, it could be trailblazing, said Marvi Rebueno Trudeau, head of the public-private partnership, including the Shell Foundation and the Global Fund, that is testing the technology. ‘The way we’re doing it right now, the microscopers collect the blood, make a slide, put it in the microscope. They then record it in the record books and at the end of the month they submit this record to the municipal health office. They then encode it and bring it to the provincial health office, and then all that information’s collated and it goes to the national level. If you do it with a smartphone, that information’s there instantly.’
Recording cases is as important as diagnosis in preventing outbreaks, because it allows the authorities to mobilise a response. ‘It’s very critical to mobilise as quickly as you can and find the equivalent of patient zero, to find out where the outbreak originated from,’ said Mario Domingo, founder of lloopp, the company that is developing the algorithm. ‘Once that’s done, we can trigger notifications through emails, SMS and apps to specific government agencies and NGOs. Sometimes the action is to mobilise doctors or it could be to trigger an order for mosquito nets, so mobilising medical teams and also the supply chains.’
The way it works is that Android smartphones are fitted with a standard 100x magnification lens. If someone goes to a rural health centre with symptoms, a volunteer can take a blood sample and, instead of examining it themselves through a microscope, simply take a picture with the phone. The algorithm then compares this picture with its database of some 50,000 blood samples, and identifies whether malaria is present.
‘It wasn’t a medical breakthrough at all,’ said Domingo. ‘We were looking for Nike swooshes and the three stripes of Adidas on Instagram and SnapChat. To do customer analytics on social media, we use machine learning to try to recognise patterns in images and see if people are getting ready to buy. We were trying to predict behaviour and I said, you know what, something a lot easier than this is blood cells.’
Registering the diagnoses electronically also helps track down the source of the outbreak. You can see which village an infected person has come from, or whether all the infected people are in the same family or living in the same house. Said Trudeau, ‘You’ll be able to figure out very quickly why there’s a case in that area. That’s the beauty of it.’
- More like this: The anti-viral app
- More like this: Better healthcare for less
- More like this: How to save lives with a nudge
This technology could be especially important because malaria rates are actually in steep decline. The number of cases in the Philippines has dropped by more than 75% since the year 2000, in large part because people in hundreds of remote villages were trained to use microscopes to make diagnoses. Last year the Philippines recorded some 4,900 cases, as opposed to 86,000 back in 1990. But as the number of cases decreases, the volunteer microscopers get out of practice, lose their skills or simply move on. The smartphone promises gold standard diagnosis at the touch of a button.
Nor is the inexorable drive towards eradication limited to the Philippines. The global incidence of malaria dropped 37% between 2000 and 2015, according the World Health Organisation. This year, a highly symbolic victory was won when Sri Lanka declared itself malaria-free. The island came within a whisker of eradication in the early 1960s – recording only 17 cases in 1963 – but its anti-malaria program fell apart, and the number of cases rose back to 250,000 in 1998.
Dozens of other countries may eradicate malaria by 2020. The WHO lists 13, including Argentina and Turkey, that have recorded no cases for more than a year, and another 21, including China, Malaysia and Iran, that could be free of the disease by the end of the decade.
Artificial intelligence of this kind being tested in the Philippines might provide the straw that breaks malaria’s back – but it isn’t fully working yet. In the latest tests, the algorithm was 85% accurate, still short of the 95% needed to equal the manual tests currently in use. That is in part because the project has been feeding the algorithm the lowest-quality slides, which are hardest to read, but which the algorithm will have to learn to read if it is to fulfil its promise.
If it succeeds, however, the potential goes far beyond malaria. Said Domingo, ‘The roadmap now includes infectious diseases like Zika, MERS and dengue fever. The process is exactly the same. You just need to arm the algorithm.’
(Pictures via Flickr: Armed Forces Pest Management Board; Department of Foreign Affairs and Trade, and The Global Fund)