MOLT: An Internet Game for Diagnosing Malaria
Twenty-four small images of red blood cells appear on the screen. Your job is to click on any that have a malarial parasite inside, removing the image. When you’ve removed all the infected cells, click on “Label all Negative” and another twenty-four cells appear. At the end, you’ll get a score and some information about how many correct choices you made.
The game is called MOLT, and it was designed by the Ozcan Research Group at UCLA. Anyone can register and play. The idea is that anyone can be given some basic information about what malarial parasites look like in red blood cells and then be part of an accurate means of correctly diagnosing the disease without having to rely on experts in the field. This would be a huge improvement for malaria diagnosis in parts of the world where malaria kills millions each year and people skilled in diagnosis are rare.
A pilot study of the game using 20 gamers produced results that were within 1.25% of the accuracy of actual experts adept at recognizing malaria, which is pretty impressive. One can imagine an arrangement where someone puts a blood film on a microscope somewhere in Asia or Africa, the images are sent out electronically to potentially millions of gamers around the world, and the answer comes back, positive or negative, in a very short time. If the pilot is any indication, the answer would agree, most of the time, with what an expert would have said.
This has implications for lots of other things that are done by microscopy or other types of imagery: pap smears, fecal smears for parasites, pathology slides etc. It could be improved upon by adding automated scanning techniques and actual experts to the crowd of gamers. These things, plus a larger number of gamers would likely be even more accurate than the gamers used in the pilot. It’s exciting.
I’ve played the game – a number of times. I have lots of experience with reading blood films for malaria, and my biggest issue with the game is that the resolution – the sharpness – of the images is often not good enough
for me to feel completely comfortable with my choices. Platelets sitting on top of red blood cells can look like a parasite. So can debris on the slide. A red cell that’s damaged, or crunched up against another cell, or too darkly stained, or abnormal in some way, etc. etc., doesn’t look like it should to begin with.
I always want to look around a bit, see what the rest of the slide looks like, look for those particular features of a malarial parasite that leave no doubt. In other words, I have a very difficult time deciding whether something is positive or negative on the basis of only one cell (unless the resolution is very good).
My other complaint is with the scoring. I find it ambiguous. When they say “Correct Positive Diagnosis 91%” does that mean 91% of the cells marked as positive were actually positive (false positives), or 91% of positive cases were identified (false negatives). I think it means the latter, because it seems the fussier I am about calling something positive, the more my positive diagnosis score goes down. For anyone trying to improve at the game, clarification on this is important.
Of course I understand that the point is that people who are not experts, and not demanding in terms of excellent microscopic optics and parasite features, can still get the right answer if there are enough people providing input. From that perspective, I think the game is brilliant, and I hope it changes the world.
Play the game on Biogames
Read the paper:
Mavandadi S, Dimitrov S, Feng S, Yu F, Sikora U, et al. (2012) Distributed Medical Image Analysis and Diagnosis through Crowd-Sourced Games: A Malaria Case Study. PLoS ONE 7(5): e37245. doi:10.1371/journal.pone.0037245