This is really important. I caught Malaria in Ghana, but the symptoms did not develop until I was in the UK. The hospital wanted to identify the type of Malaria I had before treating me, which I think was a mistake. In Ghana they take a blood sample and then immediately start treatment. The treatment can always be altered if the strand is different from the best course of action.
But the issue was they had a hard time identifying the type because I did not have one type of Malaria but two. In a hospital in the temperate world they do not have the cases to gather the skills to be able to quickly identify the strand.
I co-founded a startup with Mark Plaskow earlier this year to help make hematological analysis including detection of parasites more accurate and affordable. Mark has been doing image based detection of parasites for years and is published here: https://malariajournal.biomedcentral.com/articles/10.1186/14...
There are various issues with capturing images of this quality level in the field. It’s something we have made a lot of progress on.
We hope our devices can have an outsized impact in these regions.
If anyone reading is interested in our work we have a teaser page live at http://www.paean.io and can be reached by email using info@paean.io
Thanks for posting, I've been looking into this space. Have you looked at the College of American Pathologists Blood Parasites survey? The organisms one may encounter in the peripheral smear of a patient with fevers in the tropics is not limited to malaria (filariasis, babesiosis, etc). How are you dealing with those?
This is the kind of thing I'm expecting to see more of coming from the ML field. Orgs fighting malaria should take these efforts into consideration, back them, develop ever more robust tools. Which might already be happening, I don't know.
> I found a great dataset that consists of 27,558 single cell images with an equal number of infected and uninfected cells.
This seems to be the key of the article. I wonder how the author would have proceeded without this data-set and if they would have been able to build anything useful at all without it.
Great work, you might consider not just feedback from experts, but also from the patients themselves, perhaps if they have their own cell phone they can report feeling sick, or testify they did not get sick etc...
But the issue was they had a hard time identifying the type because I did not have one type of Malaria but two. In a hospital in the temperate world they do not have the cases to gather the skills to be able to quickly identify the strand.