NIR and hyperspectral imaging in the diagnostic of the snails intermediate hosts of Schistosomiasis in Brazil

Schistosomiasis is an important snail transmitted parasitic disease in tropical and subtropical regions. In Brazil, where 6.4 million people in 18 states are affected by this disease, three species of freshwater snails act as vectors: Biomphalaria glabrata, B.tenagophila and B.straminea. The differential diagnostic of these species is done through the shell and anatomical characteristics. Diagnostic characters are mainly in the male and female reproductive system organs, mantle and ovotestis. The calcareous shells, despite the good preservation, have very similar characteristics among the species. In order to improve the diagnosis of these species of medical importance, using only the shell, considering that many times at Museum Collections and even in the field, only the shells are available, this project aims to develop a spectroscopic method using near infrared spectroscopy and hyperspectral imaging (HI-NIR) to classify the shells of three species of snails using chemometrics tools. Twenty-eight shells of each species (B.glabrata, B.tenagophila and B.straminea) were used to obtain the spectra. The shells were provided by the Laboratório de Malacologia/ IOC/Fiocruz. The snails were anesthetized with menthol, immersed in water at 70o C, and with the aid of forceps, the shells were separated to visceral part. The shells were maintained in water for three days, and then were dried on absorbent paper. The spectra were obtained using the NIR and hyperspectral imaging system (Sisuchema, Specim). The left side of the shells were scanned using 36 µm of spatial resolution and 10 nm of nominal spectral resolution, by focusing first at the umbilicus and after at the last lap. The average spectrum of umbilicus region only (the origin of the shell) and the average spectrum of the last lap only (the freshest part of the shell) were obtained for all shells, in order to have information from these two taxonomic characters. The whole spectra (928-2554 nm) were pre-processed using SNV (standard normal variate) and the data were submitted to PCA (Principal Component Analysis) and SIMCA (Soft Independent Modeling of Class Analogy). The PCA analysis done with all sample set (Nt=168 samples: composed by 3 species, 28 shells from each species and 2 taxonomic characters of each shell) showed that the two taxonomic characters were well separated by the just 2 PC. Indeed, for all the species, there was a considerable difference between the umbilicus region spectra and the last tap spectra, especially at 1800-2100 nm. When applying PCA to the umbilicus spectra set (Nu=84) and to the last lap spectra set (Nl=84) separately, a better separation of the three species was achieved using the umbilicus spectra set. This result suggests that the usage of one taxonomical character may give better results for species identification than the usage of the entire shell. This evaluation is only possible because of imaging analysis. Using Nt=168, the SIMCA model was applied and presented an average correct prediction rate of 95%.