Automating Triatomine Identification
This week we published a new paper, "Automated identification of insect vectors of Chagas disease in Brazil and Mexico: the Virtual Vector Lab", in PeerJ. This is the first of (hopefully) many products to come out of the Virtual Vector Lab project, which started at the University of Kansas when I was in the last year of my PhD program. The aim of the project is to facilitate the identification of potential insect disease vectors to species by public health workers by uniting technology with taxonomic expertise. The project is interdisciplinary, involving biologists, computer scientists, and artists at institutions in the US, Brazil, and Mexico.
My role in the project was to read through the taxonomic literature on triatomine insects (aka assassin bugs, the vectors of Chagas disease) and identify potential morphological landmarks from which we could obtain measurements to identify specimens. I then performed the measurements by hand using digital images of museum specimens and developed a preliminary classification algorithm for identifying the specimens. Based on this preliminary work and in collaboration with an expert on image recognition software, I then optimized the landmarks for identification power and automatability of measurement collection.
My collaborators were further able to further improve the ability of our workflow to automatically identify insects by generating a candidate species lists based on distribution maps of all triatomine vectors and the location from which the image was collected (this information is available as a matter of course when an image is taken with a smartphone, all of which have built-in GPS). When all was said and done, we had an automated identification success rate of over 80%. Not bad for a computer!
To further improve our success rate in the future, we are working to facilitate the collection of standardized through a 3D printable smartphone stand that will provide a standard background color, lighting, and distance from the subject.
It's another exciting day in the life of a museum scientist!
Gurgel-Gonçalves, R., Komp, E., Campbell, L.P., Khalighifar, A., Mellenbruch, J., Mendonça, V.J., Owens, H.L., de la Cruz Felix, K., Peterson, A.T. and Ramsey, J.M., 2017. Automated identification of insect vectors of Chagas disease in Brazil and Mexico: the Virtual Vector Lab. PeerJ, 5, p.e3040.
Open access article:
It’s about time—the New World Swallowtail Butterfly project has another expedition up. This is the last batch of McGuire Center specimen images I am collecting for a study on the relationships between morphological variation and geography. This collection provides an excellent record of morphological variation across the distributions of these species.
You may also come across some specimens that look different from the other McGuire Center specimens—their backgrounds are white foam with a white ruler for scale. These images were generously provided from the private collections of dedicated amateur lepidopterists. The specimens come from a hybrid zone between two species--Papilio canadensis, Canadian tiger swallowtail, and Papilio glaucus, Eastern tiger swallowtail. We are interested in understanding whether the hybrid species looks more or less like one of its parent species, an amalgamation of the two, or if it has begun to display morphological characteristics that are completely unique.
As with the previous Swallowtail expedition, remember that there are two images for each specimen—a front and a back. This is important, because in some cases, the labels in the image have different data written on each side. Thanks for your help, and look closely—some of these specimens provide a unique historical record of biodiversity that has since been lost!
Check back when the expedition is complete—we’ll have some exciting preliminary data for you!
--Originally posted on Notes From Nature blog, 4 April, 2017
|Hannah L. Owens||
What's Going On?
Universitetsparken 15, byg 3
2100 Copenhagen Ø, Denmark
Copyright © 2015