PlantSeg

PlantSeg performs instance-aware segmentation of individual cells in densely packed three-dimensional (3D) volumetric images of plant tissues to enable quantitative analysis of morphogenesis.

Key Features:

• Two-stage segmentation strategy: Combines a convolutional neural network (CNN) to predict cell boundaries with graph partitioning algorithms to produce instance-aware cell segmentations in 3D volumetric images.
• Versatility across imaging modalities: Trained on datasets from fixed and live plant organs imaged with confocal and light sheet microscopy, enabling segmentation across varied tissues, scales, and acquisition settings.
• Application in developmental biology: Provides cell-level segmentation outputs used to study morphogenetic processes in developing plant tissues.

Scientific Applications:

• Quantitative morphogenesis analysis: Enables quantitative analysis of how individual cells contribute to tissue-level morphogenesis through high-resolution instance segmentation.
• Comparative studies across species or stages: Supports comparative analyses across species or developmental stages by generalizing across different tissues and imaging conditions.

Methodology:

A CNN predicts cell boundaries in 3D images and graph partitioning algorithms segment cells based on those boundary predictions; models were trained on confocal and light sheet microscopy datasets from fixed and live plant organs.