Automatic detection of the Cyprideis torosa (Jones, 1856) sieve pores from backscattered Electron Scanning Electron Microscopy images and development of morphometric tools for their shape identification

Abstract

Abstract Round-shaped sieve pores of Cyprideis torosa have negative correlation with salinity and are a useful proxy for reconstructing paleo-salinity trends, especially for oligohaline to mesohaline waters. However, given its time-consuming character, this method has only been used rarely. A protocol for the automatic detection of sieve pores from Backscattered Electron Scanning Electron Microscopy (SEM-BSE) images and ImageJ-FIJI software has been developed. The use of SEM-BSE images has optimized the sieve pore contrasts and extend the observation possibilities of sieve pores on less well-preserved valves. Automatic detection significantly reduces analysis times, by avoiding individual pore measurements and applying batch processes on samples. Rosenfeld and Vesper (1977) proposed an elongation index (length/beam) to discriminate round and elongated shapes, but the identification of irregular shapes only depends on the operator's appreciation. Due to the rapid acquisition of morphometric data with the automatic detection method, it is possible to systematize the use of metrical tools to discriminate sieve pore shapes. Two methods were developed from a repository based on metric variables obtained by manually digitalized 1490 sieve pores: (i) a Functional Discriminant analysis based on 5 variables (Feret's aspect ratio, roundness, aspect ratio and two ratios: Apore/AMinimum Bounding Circle and Ppore/PMinimum Bounding Circle) and (ii) an irregularity index (Roundness/(Apore/AMinimum Bounding Circle), as a complement to the elongation index. In addition, a bootstrap on 1774 sieve pores indicates that is useful to analyze over 10 valves per sample in order to estimate the proportions of these sieve pore shapes with an acceptable level of precision.