CellSeg

CellSeg performs nucleus segmentation and signal quantification in highly multiplexed fluorescence tissue images to enable single-cell analyses and improve immune cell population resolution.

Key Features:

• Pre-trained Mask R-CNN: Uses a pre-trained Mask R-CNN architecture for nucleus segmentation in fluorescence images, obviating the need for additional model training or manual dataset annotation.
• Signal Quantification: Performs signal quantification across multiplexed fluorescence channels to support downstream single-cell analysis.
• Performance and Generalization: Matched or surpassed top segmentation algorithms in the 2018 Kaggle Data Challenge and generalizes across diverse multiplexed cancer tissue images.
• Automated Post-Processing: Includes automated post-processing steps that enhance the resolution of immune cell populations for more precise downstream single-cell analyses.
• Integration with Imaging Pipelines: Applied to a highly multiplexed colorectal cancer dataset acquired via CO-Detection by indEXing (CODEX), demonstrating integration into tissue imaging workflows and identification/validation of cell populations.

Scientific Applications:

• Cancer Tissue Imaging: Enables detailed analysis of tumor microenvironments through precise nucleus segmentation and signal quantification in multiplexed tissue images.
• Immunology and Oncology: Improves resolution and identification of immune cell populations to support immunology- and oncology-focused single-cell studies.
• Multiplexed Dataset Validation: Facilitates identification and validation of cell populations in multiplexed imaging datasets such as CODEX-acquired colorectal cancer samples.

Methodology:

Segmentation is performed by a pre-trained Mask R-CNN model followed by automated post-processing steps to enhance immune cell resolution, with signal quantification across multiplexed fluorescence channels (demonstrated on CODEX colorectal cancer data).