TOWARDS UNDERSTANDING SPECIATION BY AUTOMATED EXTRACTION AND DESCRIPTION OF 3D FORAMINIFERA STACKS

The sheer volume of 3D data restricts understanding of genetic speciation when analyzing specimens of planktonic foraminifera and so we develop an end-to-end computer vision system to solve and extend this. The observed fossils are planktonic foraminifera which are single-celled organisms that live in vast numbers in the world's oceans. Each foram retains a complete record of its size and shape at each stage along its journey through life. In this study, a variety of individual foraminifera are analyzed to study the differences among them and compared with manually labelled ground truth. This is an approach which (i) automatically reconstructs individual chambers for each specimen from image sequences, (ii) uses a shape signature to describe different types of species. The automated analysis by computer vision gives insight that was hitherto unavailable in biological analysis: analyzing shape implies understanding spatial arrangement and this is new to the biological analysis of these specimens. By processing datasets of 3D samples containing 9GB of points, we show that speciation can indeed now be analyzed and that automated analysis from morphological features leads to new insight into the origins of life.