Skip to content

Monai training of an MS lesion segmentation model #12

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Draft
wants to merge 108 commits into
base: main
Choose a base branch
from
Draft

Monai training of an MS lesion segmentation model #12

wants to merge 108 commits into from

Conversation

plbenveniste
Copy link
Collaborator

This PR details the work to build a model using monai to automatically segment MS lesions in the SC.

It contains the following files:

  • 1_create_msd_data.py : to create an MSD dataset, from the multiple MS lesion dataset, usable by monai
  • config.yml : the config YAML file used to configure the parameters of the model and the training
  • losses.py : where we define the losses used during model training
  • requirements.txt : where we list the libraries to install using pip
  • train_monai_unet_lightning_multichannel_input.py : the main script to train a model with multiple input channel (image and SC seg) and one output (lesion seg)
  • train_monai_unet_lightning_multichannel_output.py : the main script to train a model with one input (the image) and multiple outputs (SC seg and lesion seg)
  • train_monai_unet_lightning.py : the main script to train a model with one input (image) and one output (lesion seg)
  • utils.py : a file which contains other useful function for model training

This is a work in progress.

@plbenveniste
created script to build msd dataset for monai nnunet model training
bbde8f2
@plbenveniste
added requirements script
081276a
@plbenveniste
removed yaml from requirements
0ef3f0a
@plbenveniste
changed output file names
451ca30
@plbenveniste
added missing requirements
1dbd553
@plbenveniste
initialised config.yml file example
91bac32
@plbenveniste
initialised main file
48f109c
@plbenveniste
initialised models file
5aa2eae
@plbenveniste
initalised transforms file
5b233d2
@plbenveniste
modified to fit our training parameters
794430a
@plbenveniste
simplied main training script
0619980
@plbenveniste
fixed canproco problem (having both img and label with same link)
e04c423
@plbenveniste
added training script for a monai trained UNETR
26629c8
@plbenveniste
added fake config file (for debugging)
e815e14
@plbenveniste
updated requirements to add monai[all] for problem with data loading
5aaf6db
@plbenveniste
removed old config file
8eafeff
@plbenveniste
changed links to dataset
8f43eac
@plbenveniste
fixed training. Still need to fix validation params
86e1a32
@plbenveniste
removed files using pytorch ligthning training
32ca1cd
@plbenveniste
working monai script based on Jan's code: but no dice score improvement
5647955
@plbenveniste
pytorch lightning script based on Naga's work
a4d7958
@plbenveniste
modified requirements for pytorch lightning
4e768e4
@plbenveniste
added multiply by -1 transform
186c602
@plbenveniste
parameters changed for config file for first inference run
adaf04f
@plbenveniste
added SoftDiceLoss
dab7dcc
@plbenveniste
removed print from inverse function in utils
80494e5
@plbenveniste
changed resolution to 0.6 isotropic
4727dbc
@plbenveniste
added plot images function for wandb
5992732
@plbenveniste
changed loss function and added image printing
9aefb6d
@plbenveniste
changed some parameters for training
855e26f
@plbenveniste
added contrast, site and orientation in msd dataset
88b6619
@plbenveniste
improved computation of orientation of image
f387067
@plbenveniste
added __init__.py file for import possibility
ec256ab
@plbenveniste
removed unused files
554b522
@plbenveniste
moved files to utils folder
1fdbdd3
@plbenveniste
updated parameters for model testing
8292ff8
@plbenveniste
updated inference script and evaluation plots scripts
7e361ba
@plbenveniste
added removal of .nii.gz for UINT1 contrast
bf07262
@plbenveniste
changed workers to 0 for test_model
9f5587a
@plbenveniste
added more info in output
c23d63f
@plbenveniste
created file for cropping aroung head
034d97e
@plbenveniste
updated training script to sota model training script (set workers to 0)
00ec30c
@plbenveniste
changed location of saving of yaml file to save with the same date as…
e038c67 
… json file
@plbenveniste
init mednext training script
391193c
@plbenveniste
added library for diffusion model
25e7874
@plbenveniste
first draft (non-functionning) of diffusion model training script
b7ee720
@plbenveniste
created script to train a mednext model
6f807f9
@plbenveniste
removed cropping of image before inference
74b6eed
@plbenveniste
added new config files
c497c4e
@plbenveniste
added script to perform inference and compute the dice score with var…
825d816 
…ying threshold
@plbenveniste
fixed .cpu problem and added more thresholds
ba6d90e
@plbenveniste
first draft of script for TTA
9027f32
@plbenveniste
added code to plot the opt threshold output
11eec85
@plbenveniste
fixed threshold to 0.5
38c1594
@plbenveniste
fixed parenthesis when computing dice score
e3b1eff
@plbenveniste
added script to compute TTA with 2nd strategy
a37c6a0
@plbenveniste
added script for mednext inference
60506b4
@plbenveniste
added computation of f1-score, ppv and sensitivity
a79cb44
@plbenveniste
fixed utils command
953de3f
@plbenveniste
removed dataset aggregation scripts
15d13b4
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@plbenveniste