ultrasound_simulator_gan_model_zoo.md

Freehand Ultrasound Image Simulation with Spatially-Conditioned Generative Adversarial Networks

This page describes how to acquire and use the network described in

Yipeng Hu, Eli Gibson, Li-Lin Lee, Weidi Xie, Dean C. Barratt, Tom Vercauteren, J. Alison Noble (2017). Freehand Ultrasound Image Simulation with Spatially-Conditioned Generative Adversarial Networks, In MICCAI RAMBO 2017

Downloading model zoo file and conditioning data

If you cloned the NiftyNet repository, the network weights and examples data can be downloaded with the command

net_download ultrasound_simulator_gan_model_zoo

(Replace net_download with python net_download.py if you cloned the NiftyNet repository.)

Alternatively, you can manually download:

• model zoo entry
• example data
• trained network weights

and unzip:

• ultrasound_simulator_gan_code.tar.gz into ~/niftynet/extensions/ultrasound_simulator_gan/
• ultrasound_simulator_gan_model_zoo_data.tar.gz into ~/niftynet/data/ultrasound_simulator_gan/
• ultrasound_simulator_gan_weights.tar.gz into ~/niftynet/models/ultrasound_simulator_gan/

Make sure that the model directory (~/niftynet/extensions/ by default) is on the PYTHONPATH.

This network generates ultrasound images conditioned by a coordinate map. Some example coordinate maps are included in the model zoo data. Additional examples are available here).

Generating segmentations for example data

Generate segmentations for the included example conditioning data with the command

net_gan inference -c ~/niftynet/extensions/ultrasound_simulator_gan/config.ini

Replace net_segment with python net_gan.py if you cloned the NiftyNet repository.

Replace ~/niftynet/ if you specified a custom download path in the net_download command.

Generating segmentations for additional conditioning data

Editing the configuration file

Make a copy of the configuration file ~/niftynet/extensions/ultrasound_simulator_gan/config.ini to a location of your choice. You may need to change the path_to_search and filename_contains lines in the configuration file to point to the correct paths for your conditioning data. You can also change the save_seg_dir setting to change where the segmentations are saved.

Generating samples

Generate samples from the simulator with the command net_gan.py inference -c edited_config.ini, replacing edited_config.ini with the path to the new configuration file. Sets of simulated US images interpolated between two samples will be generated in the path specified by the save_seg_dir setting with names of the form k_id_niftynet_generated.nii.gz, where k is the interpolation index 0-9 and id is the frame code from the input conditioning data filename.