Identifying Chagas disease vectors using Elliptic Fourier Descriptors of body contour: A case for the cryptic dimidiata complex

Abstract Background Triatoma dimidiata (Reduviidae: Triatominae) is an important vector of Chagas disease in various countries in the Americas. Phylogenetic studies have defined three lineages in Mexico and part of Central America, but methods for identifying them using morphometric analyses with landmarks have not yet been resolved. Elliptical Fourier Descriptors (EFDs), which mathematically describe the shape of any closed two-dimensional contour, could be a potentially useful alternative method. The objective of this work was to validate the use of EFDs for the identification of the three lineages of this species complex.Method A total of 84 dorsal view photographs of individuals of the three lineages were used. The body contours were described with EFDs using between 5 and 30 harmonics. The number of obtained coefficients was reduced by a Principal Components Analysis and the first axis scores were used as shape variables. A Canonical Variances Analysis, a linear discriminant function analysis and a multilayer perceptron neural networks were then performed using the shape variables, identifying the minimum number of harmonics sufficient to produce efficiently classifications.Results The first principal component explained 50% of the variability, regardless of the number of harmonics used, but the results of both, Principal Component Analysis and Canonical Variance Analysis get improved by increasing the number of harmonics and components considered. With 25 harmonics and 8 components, the identification of haplogroups was achieved with an overall efficiency greater than 97%. The 30 multilayer perceptron neural networks were also efficient in identification, reaching 91% efficiency with the validation data.Conclusions The use of EFDs is a simple and useful method for the identification of major lineages of Triatoma dimidiata. This method outperform other novel approaches, therefore could serve as an automated identification method.