254-9 Bites and Bytes: Investigating Echinoid Feeding Performance Using Deep-Learning-Assisted Muscle Reconstruction

Session: Recent Advances in Fossil Imaging

Presenting Author:

Authors:

Abstract:

Owing to their relatively rich fossil record, echinoids have been a topic of substantial interest for interpreting the ecological implications of macroevolutionary changes in morphology. Principle among these has been the evolution of their feeding apparatus, the so-called Aristotle’s lantern, and its relationship to biomechanical performance for various feeding modes. However, due to typical non-preservation of soft tissues, making these interpretations in the fossil record relies on the morphology of the biomineralized test. In the case of feeding, the perignathic girdle, which supports the muscular articulations of Aristotle’s lantern, is often preserved in the fossil record and could be used to provide insight into deep-time feeding performance. Two large muscle groups attach to this structure and both their size and arrangement may be related to feeding mode. This information may be possible to estimate for fossil taxa based on girdle morphology, but first requires a robust understanding of the relationship of musculature to perignathic girdle morphology in the extant taxa. Here, we present our process in developing this foundation. We discuss a novel workflow leveraging deep learning tools, Python automations, and high-performance computing to significantly reduce the time needed to segment over 100 three-dimensional (3D) imaging datasets and convert them to mesh-based 3D models. From these models we collect biomechanically relevant measurements such as muscular volume, cross-sectional area, and attachment surface area and relate these to the geometry of the girdle. We find that a diversity of perignathic girdle and muscle arrangements are present across the “regular” Echinoidea, and that distinct anatomies are associated with particular feeding modes and life habits. The benefit of this work is two-fold. This dataset can serve future interpretation of echinoid macroevolutionary trends as well as more robust biomechanical investigations of these marine organisms. Additionally, the scan-to-model pipeline we have developed is system-agnostic and scalable, which will enable the development of similarly large datasets in the future, thus bolstering future functional morphology research.

Geological Society of America Abstracts with Program. Vol. 57, No. 6, 2025

doi: 10.1130/abs/2025AM-10779

© Copyright 2025 The Geological Society of America (GSA), all rights reserved.