The field of comparative morphology has been revolutionized by the application of high resolution digital imaging methods to non-invasively visualize complex physiological features. However, truly quantitative characterization and comparison of complex morphological features still cannot be adequately addressed by existing methods, which are typically developed only for idealized 2D images or surfaces. The accurate and efficient characterization and comparison of shapes with 3D noisy imaging data represents highly non-trivial computational problems which have yet to be adequately addressed in comparative morphology studies, nor has efficient computational software been made available to the researchers. The goal of this project is to develop advanced computational methods for accurate quantitative characterization and comparison of specimen morphology from high resolution 3D voxel-based digital imaging modalities. With the growing recognition of the importance of digital libraries of biological specimens derived from advanced imaging methods, such as the NSF funded Digital Fish Library (DFL) and Digital Morphology (DigiMorph) projects, advanced methods for utilizing these data are of great importance, but pose significant technical challenges. Two broad classes of problems are of critical importance: 1) The ability to quantitatively characterize complicated morphological features from high resolution volumetric data and 2) Methods for comparing such features between specimens. Our objective is to develop two specific computational methods for geometric morphological analysis that can optimally characterize and compare geometric features embedded within real 3D imaging data, and are robust to noise and resolution limitations: 1) A novel shape analysis method based on signatures derived from spherical wave decomposition of 3D images; 2) A robust non-linear spatial normalization method based on diffeomorphic image and landmark registration. The spatial normalization methods will allow homologous structures to be correctly non-linearly warped to each other or a common template for comparison, while the decomposition method will facilitate robust, efficient, accurate, and automated characterization of shapes embedded within complex 3D datasets. These methods can then be used to generate species-specific atlases that define normative morphologies, thus facilitating both inter- and intra-specific comparative analyses. These methods will then be applied to the general problem of automated shape segmentation, then tested on two problems of significant biological importance: 1) Co-evolution of the short-tailed opossum inner ear and cranium and 2) Three-spine stickleback evolution from freshwater to saltwater species.
Characterization and comparison of morphological (form, shape, or structure) variations is a problem of significant impact across a wide range of biological disciplines. The rise of 3D volumetric imaging methods for digitizing biological samples offers great possibilities for addressing these issues but requires a theoretical and computational framework capable of allowing researchers efficient and accurate methods for analyzing complicated biological structures embedded within 3D volumetric noisy digital data. The goal of this project is to develop computational tools to address the two primary issues at the heart of these analyses: The ability to accurately and efficiently 1) characterize complex morphological features and 2) compare morphological features between specimens. The ability to perform these is critical to facilitating the use of all digital library data for quantitative morphology but to date have not been developed. The goal of this proposal is to develop analysis software to fill this significant gap in the bridge between digital imaging methods and its ultimate potential for transforming the field of comparative morphology by developing computational methods to address a broad range of morphological questions that inform our knowledge of the evolution and diversification of species. The methods developed by this project will greatly extend the capabilities of researchers and students to incorporate quantitative anatomical measurements into the study of evolutionary biology. The methods are general and applicable to any 3D imaging modality and thus will be of utility to any digital library and will serve as a platform on which new technologies and methodologies can be applied in the future. The resulting analysis tools will be open source and disseminated to researchers through the DFL website (www.digitalfishlibrary.org). A corresponding public education exhibit will be developed at the Cabrillo Marine Aquarium (www.cabrillomarineaquarium.org/). Developing a general computational platform for the computational morphology from 3D digital data will allow evolutionary biologists to quantitatively and reproducibly address problems that provide greater insight into how ecological parameters might be, quite literally, 'shaping' biodiversity and thus has potentially profound implications for the field of evolutionary biology.
