Nucleus Segmenter

Overview

This model detects nuclei in a microscopy image and specifies the pixels in the image that are assigned to each nucleus. The model is developed based on the architecture of Mask R-CNN using Feature Pyramid network (FPN) and a ResNet50 backbone. Given an image (of size 64 x 64, 128 x 128 or 256 x 256), this model outputs the segmentation masks and probabilities for each detected nucleus. The mask is compressed using Run-length encoding (RLE).

The model is based on the TF implementation of Mask R-CNN. The model is trained on the Broad Bioimage Benchmark Collection (Accession number BBBC038, Version 1) dataset of annotated biological images.

Model Metadata

References

• He, K., Gkioxari, G., Dollár, P. and Girshick, R., 2017, October. Mask R-CNN. In Computer Vision (ICCV), 2017 IEEE International Conference on (pp. 2980-2988). IEEE.
• Ljosa, V., Sokolnicki, K.L. and Carpenter, A.E., 2012. Annotated high-throughput microscopy image sets for validation. Nature methods, 9(7), pp.637-637.
• Broad Bioimage Benchmark Collection [Ljosa et al., Nature Methods, 2012].
• Mask R-CNN Github Repository

Licenses

Options available for deploying this model

This model can be deployed using the following mechanisms:

• Deploy from Dockerhub:

docker run -it -p 5000:5000 codait/max-nucleus-segmenter

• Deploy on Kubernetes:

kubectl apply -f https://raw.githubusercontent.com/IBM/MAX-Nucleus-Segmenter/master/max-nucleus-segmenter.yaml

• Locally: follow the instructions in the model README on GitHub

Example Usage

You can test or use this model

• using cURL

Test the model using cURL

Once deployed, you can test the model from the command line. For example:

$ curl -F "image=@samples/example.png" -XPOST http://localhost:5000/model/predict

You should see a JSON response like that below:

{
  "status": "ok",
  "predictions": [
    {
      "mask": [
        3507,
        1,
        3571,
        5,
        3635,
        6,
        3700,
        5,
        3766,
        4,
        3831,
        2
      ],
      "probability": 0.9837305545806885
    }
    ...
  ]
}