Voje, K. L. | bioRxiv Science

bioRxiv2022

DeepBryo: a web app for AI-assisted morphometric characterization of cheilostome bryozoans

O_LIBryozoans are becoming an increasingly popular study system in macroevolutionary, ecological, and paleobiological research. Members of this colonial invertebrate phylum are notable for displaying an exceptional degree of division of labor in the form of specialized modules (polymorphs), which allow for the inference of individual allocation of resources to reproduction, defense, and growth using simple morphometric tools. However, morphometric characterizations of bryozoans are notoriously labored, due to the high number of structures often captured per image, as well as the need for specialized knowledge necessary for classifying individual skeletal structures within those images. C_LIO_LIWe here introduce DeepBryo, a web application for deep learning-based morphometric characterization of cheilostome bryozoans. DeepBryo requires a single image as input and performs measurements automatically using instance segmentation algorithms. DeepBryo is capable of detecting objects belonging to six classes and outputting fourteen morphological shape measurements for each object based on the inferred segmentation maps. The users can visualize the predictions, check for errors, and directly filter model outputs on the web browser. Measurements can then be downloaded as a comma-separated values file. C_LIO_LIDeepBryo has been trained and validated on a total of 72,412 structures, belonging to six different object classes in 935 SEM images of cheilostome bryozoans belonging to 109 different families. The model shows high (>0.8) recall and precision for zooid-level structures. Its misclassification rate is low (~4%) and largely concentrated in a single object class (opesia). The models estimated structure-level area, height, and width measurements are statistically indistinguishable from those obtained via manual annotation (r2 varying from 0.89 to 0.98) and show no detectable bias. DeepBryo reduces the person-hours required for characterizing the zooids in individual colonies to less than 1% of the time required for manual annotation at no significant loss of measurement accuracy. C_LIO_LIOur results indicate that DeepBryo enables cost-, labor,- and time-efficient morphometric characterization of cheilostome bryozoans. DeepBryo can greatly increase the scale of macroevolutionary, ecological, taxonomic, and paleobiological analyses, as well as the accessibility of deep learning tools for this emerging model system. Finally, DeepBryo provides the building blocks necessary for adapting the current application to other study groups. C_LI

bioinformatics↗

bioRxiv2019

ML-morph: A fast, accurate and general approach for automated detection and landmarking of biological structures in images

O_LIMorphometrics has become an indispensable component of the statistical analysis of size and shape variation in biological structures. Morphometric data has traditionally been gathered through low-throughput manual landmark annotation, which represents a significant bottleneck for morphometric-based phenomics. Here we propose a machine-learning-based high-throughput pipeline to collect high-dimensional morphometric data in images of semi rigid biological structures.\nC_LIO_LIThe proposed framework has four main strengths. First, it allows for dense phenotyping with minimal impact on specimens. Second, it presents landmarking accuracy comparable to manual annotators, when applied to standardized datasets. Third, it performs data collection at speeds several orders of magnitude higher than manual annotators. And finally, it is of general applicability (i.e., not tied to a specific study system).\nC_LIO_LIState-of-the-art validation procedures show that the method achieves low error levels when applied to three morphometric datasets of increasing complexity, with error varying from 0.5% to 2% of the structures length in the automated placement of landmarks. As a benchmark for the speed of the entire automated landmarking pipeline, our framework places 23 landmarks on 13,686 objects (zooids) detected in 1684 pictures of fossil bryozoans in 3.12 minutes using a personal computer.\nC_LIO_LIThe proposed machine-learning-based phenotyping pipeline can greatly increase the scale, reproducibility and speed of data collection within biological research. To aid the use of the framework, we have developed a file conversion algorithm that can be used to leverage current morphometric datasets for automation, allowing the entire procedure, from model training all the way to prediction, to be performed in a matter of hours.\nC_LI

evolutionary biology↗