tumor-tensors

Abstract : Summarize everything in a few sentences.

We built a system to convert annotated nii files into pixel contours, then fill those contours and render them into grids or mp4s.

Introduction : What's the problem? Why should we solve it?

We wanted to render both the MRI dicoms and make an augmented display of contours around labeled tumors to visually inspect how predictive models are performing.

Methods : How did we go about solving it?

We used pydom and matplot lib for rendering. We also developed some utility methods for loading nii data, converting nii labels into per-dicom/slice contours, used matplotlib and shapely to augment the dicom display, and used ffmpeg to convert plots images into an mp4 video.

Results : What did we observe? Figures are great!

MP4 Contour Rendering (unlisted YouTube video)

Conclusion/Discussion:

Please make sure you address ALL of the following:

1. What additional data would you like to have

It would have been nice to have more attribute data about segmented tumors (texture, density, etc). It also would have been nice to already have nii data in a per-dicom data structure.

2. What are the next rational steps?

Integrate these rendering tools into predictive model pipelines.

3. What additional tools or pipelines will be needed for those steps?

No extra tools are needed for rendering, it can function stand-alone.

4. What skills would additional collaborators ideally have?

Computer vision.

Reproduction: How to reproduce the findings!

Open docker or run our Jupyter notebooks.

Docker

*The Docker image contains <R/jupyter> notebooks of all analyses and the dependencies to run them. Be sure to note if you need any special credentials to access data for these analyses, don't package restricted data in your containers!

Instructions for running the following notebooks: be sure to adjust these instructions as necessary! check out https://github.com/Sage-Bionetworks/nf-hackathon-2019 for example containers and instructions

1. docker pull <your dockerhub repo>/<this container> command to pull the image from the DockerHub
2. docker run <your dockerhub repo>/<this container> Run the docker image from the master shell script

Important Resources : primary data, github repository, Synapse project, dockerfile link etc.