OpSeF

OpSeF performs deep learning-based instance segmentation of 2D and 3D bioimages to enable reproducible, modular analysis and benchmarking of segmentation models.

Key Features:

• Collaborative framework: Integrates the expertise of biomedical users and image analysts to combine domain knowledge with analytical workflow design.
• Standardized workflow design: Defines standard inputs and outputs to enable modular workflows and interoperability with other software tools.
• Semi-automated processing: Semi-automates preprocessing, convolutional neural network (CNN)-based segmentation in 2D and 3D, and post-processing.
• Model benchmarking and optimization: Supports parallel benchmarking of multiple models and optimization of pre- and post-processing parameters to reduce the need for retraining.
• Integration of CNN-based methods: Integrates U-Net (as used in Cellprofiler 3.0), StarDist, and Cellpose and supports addition of new networks.

Scientific Applications:

• Instance segmentation of bioimages: Precise segmentation of cells and subcellular structures in 2D and 3D microscopy data.
• Analysis of complex biological structures: Quantitative characterization of morphology and spatial organization in biomedical research.
• Model comparison for applied studies: Benchmarking and parameter optimization to select segmentation approaches for basic research and clinical diagnostics.

Methodology:

Computational steps explicitly include preprocessing, CNN-based segmentation in 2D and 3D using U-Net, StarDist, and Cellpose, post-processing, parallel benchmarking of multiple models, optimization of pre- and post-processing parameters, and use of standardized inputs and outputs for modular workflows.