Omnipose

Omnipose segments cells in 2D and 3D microscopy images using deep neural networks to enable precise single-cell morphological and molecular measurements.

Key Features:

• Deep neural network algorithms: Uses deep neural network algorithms for image segmentation tasks.
• Network outputs — gradient of the distance field: Produces unique network outputs, including the gradient of the distance field, to improve segmentation accuracy.
• 2D and 3D image processing: Processes both 2D and 3D images from any imaging modality when trained on representative datasets.
• Segmentation of diverse morphologies and optical characteristics: Accurately segments cells with diverse morphological and optical characteristics.
• Improved performance versus Cellpose: Outperforms Cellpose in challenging segmentation scenarios described in the input.
• Robustness to complex morphologies: Handles elongated and branched cell morphologies.
• Handling mixed and perturbed samples: Segments cells in mixed bacterial cultures and antibiotic-treated cells.
• Quantitative morphological and molecular measurement support: Enables precise and quantitative measurements of morphological and molecular phenomena across diverse cell types.

Scientific Applications:

• Mixed bacterial culture analysis: Segments individual cells within mixed bacterial cultures for single-cell analysis.
• Antibiotic response studies: Segments antibiotic-treated cells to quantify morphological changes.
• Complex morphology characterization: Characterizes elongated or branched morphologies and extreme morphological phenotypes.
• Interbacterial antagonism studies: Enables characterization of extreme morphological phenotypes emerging during interbacterial antagonism.
• Multimodal microscopy: Applies to images from varied imaging modalities and to three-dimensional objects.
• Single-cell quantitative microscopy: Supports precise single-cell morphological and molecular measurements in microscopy-based research.

Methodology:

Omnipose applies deep neural networks with network outputs that include the gradient of the distance field and is trained on representative 2D/3D imaging datasets.