Batch-Mask: Automated Image Segmentation for Organisms with Limbless or Non-Standard Body Forms

Area of Science:

• Evolutionary biology
• Bio-imaging
• Computational biology

Background:

• Color patterns are crucial for understanding evolution, sexual selection, and predator-prey dynamics.
• Automated analysis of biological color patterns is hindered by non-standard organism shapes (e.g., elongate, spiral).
• Current methods for analyzing elongate organisms rely on time-consuming manual landmarking, limiting research scope.

Purpose of the Study:

• To develop an automated, customizable workflow (Batch-Mask) for analyzing large photographic datasets of non-standard biological organisms.
• To provide an open-source solution independent of proprietary software.
• To enable efficient, large-scale quantitative analysis of color patterns in morphologically diverse species.

Main Methods:

• Developed Batch-Mask, an automated workflow for isolating non-standard biological organisms from backgrounds in photographic datasets.
• Created a user guide for fine-tuning model weights and integrating with existing tools like micaToolbox.
• Implemented a landmark-free automation approach.

Main Results:

• Batch-Mask is 60x faster than manual landmarking.
• Generated masks correctly identified 96% of snake pixels.
• Validation using micaToolbox showed no significant difference in pattern energy analysis between Batch-Mask and human segmentation.

Conclusions:

• Batch-Mask significantly reduces the time and effort required for quantitative analysis of non-standard biological subjects.
• The workflow facilitates large-scale comparative analyses of color, pattern, and shape across diverse morphologies and natural history collections.
• Landmark-free automation with Batch-Mask expands the scale and taxonomic breadth for studying color variation.