MAIA Segmentation Portal

MAIA Segmentation Portal is a user-friendly portal, hosted in MAIA, that allows users to interact with the available models, upload medical images, and receive predictions in seconds. Alternatively, users can download the models and run inference on their own machines with a single command. The models are available in the MONet Bundle format, which is compatible with MONAI Deploy and MONAI Label. The single models are deployed and made available as single MONAI Label server applications, hosted in MAIA. The MAIA Segmentation Portal is running as a KubeFlow-based project, so it is possible to run inference on the models from the KubeFlow UI. If you want to create your own MONet-based model, we provide instructions on how to run the training, starting from your annotated data, either locally or with KubeFlow, using the MONet Bundle format.

GUI Installation

To start using the MAIA Segmentation Portal, you can install the GUI application. The GUI application is available for Windows and Linux.

Download the Windows version of the MAIA Segmentation Portal from here or the Linux version from here

Installation

To install the MAIA Segmentation Portal Python API:

pip install monet-bundle

Get Access to the Portal

To access the MAIA Segmentation Portal, you need to register for an account in MAIA and request access to the maia-segmentation project in the MAIA - Sign Up Page. Once your registration is approved, you will be able to log in to the portal and start using the models. To log in to the portal, you can use the following command:

MONet_login --username <USERNAME>

You will be prompted to enter your password, and then you will be able to access the portal. If your access token is expired, you will be prompted to enter your username and password again.

List Available Models

To list the available models in the MAIA Segmentation Portal, you can use the following command:

MONet_login --username <USERNAME> --list-models

Run Remote Inference

To run remote inference on the MAIA Segmentation Portal, you can use the MONet_remote_inference command. This command allows you to upload your medical images to the portal and receive predictions in seconds.

MONet_remote_inference --model <MODEL_NAME> -i <INPUT_FILE> -o <OUTPUT_FILE>

Run Local Inference

To run local inference with the MONet Bundle, you can use the MONet_local_inference command. This command allows you to run inference on your own machine with a single command. The pre-requisites for running local inference are:

• Available GPU with NVIDIA drivers installed

• PyTorch >=2.4.1 installed

Recommended pytorch installation command:

pip install light-the-torch
ltt install torch==2.4.1 torchvision

MONet_local_inference --model <MODEL_NAME> -i <INPUT_FILE> -o <OUTPUT_FILE>