CHANGELOG.md

Changelog

Main changes to this code/ project are documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

ClinicaDL 1.6.0

Fixed

• Fix sorting particpants sessions with 2 digit,
• Fix interpret —-save_nifti issue,
• Fix BIDS file format for pet,

Changed

• Change 2 digits session label to 3 digits,
• Change black and isort for ruff and codespell,
• Update type hint and docstring,

New

• Add --fsdp option
• Add --valid_longitudinal option to allow the validation on longitudinal data,

ClinicaDL 1.6.0

Fixed

• Fix t1-volume quality-check empty tsv issue,
• Fix bug with adapt pipeline,
• Fix predict —-overwrite issue,

Changed

• Remove Clinica dependency
• Change MapsManager architecture by adding more Callbacks,
• Update python to v3.11 and pytorch to v2.0,

New

• Add missing methods in models classes,
• Add --save_all_models option to save models weights at each epochs,
• Add --skip_leak_check option to skip data leakage check,
• Add --adaptive_learning_rate option,
• Add prepare-data-from-bids pipeline to extract tensor from bids to caps,
• Add dti modality,
• Add hugging-face push/pull pipelines to share models on HuggingFace,
• Add confidence intervals and metrics

ClinicaDL 1.5.1

Fixed

• Fix retrocompatibility for new option in maps.json,

Changed

• Change MapsManager architecture by adding Callbacks,

New

• Add --fully_sharded_data_parallelism option,
• Add --emisions_calculator option with codecarbon,
• Add the semi-supervised domain adaptation network proposed for the MICCAI DART workshop.

ClinicaDL 1.5.0

Fixed

• Fix adapt command

Changed

• SSIM is now executed on GPU when possible

New

• Add new command generate artifacts to generate noise/contract/motion
• Add options for data augmentation
• Add fine-tuning option
• Add --amp for automatic mixed precision
• Add --track_exp option to track your parameters during training with MLflow or WandB

ClinicaDL 1.4.0

Fixed

• Fix --diagnoses and --merged_tsv option bug in clinicadl get-labels
• Fix Pathlib bugs
• Fix get_tsv_paths function bug.
• Fix a bug for which it was impossible to use predict without specifying the splits and selection metrics.

Changed

• Changed default batch size to 8 for clinicadl predict
• Changed VAEs main class

New

• Add --n_proc option in clinicadl generate pipelines for parallelization.
• Add --split to clinicadl predict
• Add new VAE networks.
• Add pytorch function to summarize
• Add --size_reduction and --size_reduction_factor options to clinicadl train, clinicadl predict and clinicadl interpret.
• Add SSIM2D, SSIM3D metrics for VAE.
• Add --save_latent_space option to clinicadl train and clinicadl predict.
• Add clinicadl generate trvial_motion
• Add Data augmentation with torchio

ClinicaDL 1.3.1

Fixed

• Fix TypeError when running ClinicaDL.
• Fix --extract_json option bug in clinicadl prepare-data.
• Fix clinicadl tsvtools get-labels error finding clinica iotools missing-modalities output.

Changed

• Changed clinicadl tsvtools get-labels output directory.

New

• Add --caps_directory option in clinicadl tsvtools get-labels.

ClinicaDL 1.3.0

New

• Add new command quality-check pet-linear.
• Add new command generate hypometabolic.
• Add new network architecture: Resnet3D and SqueezeExcitationCNN.
• Add flair-linear modality for prepare-data command.
• Add pytorch profiler.
• Add --save_nifti option for interpretcommand.
• Add --output_dir argument for tsvtools get-labels command

Changed

** Core: **

• Transition from os to pathlib.
• Update data CI.
• Improve maps_manager.
• Change --acq_label option for --tracer.
• Update tutorial.

ClinicaDL 1.2.0

Changed

** Core: **

• Add ClinicaDL installation with pipx.
• Improve logging.
• Add method argument to the interpret command to choose between the new Grad-CAM method and the gradient method.
• Change extract command to prepare-data.
• Change output of get-labels, split and kfold commands to one TSV per split instead of one per label.
• Change tsvtool command to tsvtools.
• Change tsvtools getlabels command to tsvtools get-labels and remove the progression column in the TSV output.
• Add new commands: tsvtools get-progression, tsvtools get-metadata, tsvtools prepare-experiment (split + kfold), and tsvtools adapt.
• Update data CI.
• Add a new model for quality check

Fixed

• Fix quality-check t1-linear

ClinicaDL 1.1.1

Changed

** Core: **

• Dependencies versions were updated in order to support Python equal or greater than 3.8.
• Security updates for some dependencies.
• Include the --save_tensor and --save_nifty options into the predict sub-command (before save_tensor was also subcommand`).
• Update some paper references in the documentation.
• Add a description for the available models.
• Add the flag --model_layers to the clinicadl train list_models subcommand to visualize the model layers.

Fixed

• Fix bug when using the clinicadl train list_models command.

ClinicaDL 1.1.0

Changed

** Core: **

• Add new VAE models for training.
• Add multiprocessing for the extract task.
• Add tests to validate resume and quality-check functionalities.
• Update minimal version for first level dependencies: clinica, scikit-learn, scikit-image.
• Add a new Makefile to the root of the repository in order to facilitate sysadmin frequent tasks. The CI script was also modified.

** Command line:**

• Update of the train command line: now all actions related to training (resume, list models, tasks…) are sub-commands of the command clinicadl train. Task is now a sub-command, and not an argument.
• Add command to list the available models for the training task and the details of each model.
• Add an option to the command save-tensor to save the reconstruction outputs as nifti files.
• Add command clinicadl train from_json to train again the same experiment from a JSON file generated by ClinicaDL.

Fixed

• Allow the user to choose the optimizer and the metric independently of the performed task.
• Logger does not display output according to the --verbose option during the training tasks. This behavior is fixed.
• Behavior of resume was changed.
• Reader to request BIDS/CAPS datasets changed its interface in the recent version of Clinica (>=0.5.4). Code using these functions was adapted.
• Exceptions raised are now more specific and detailed.
• Documentation was updated.

ClinicaDL 1.0.4

Changed

** Core: **

Switch to Poetry to manage production and development requirements. Update requirements. Improve metrics for multi-class settings. Remove NaN values in input data. Allow to choose the name of the JSON file produced by extract. Improve logs.

** Command line:** Add selection_metrics option in clinicadl train (already available in TOML file). Print help if no argument is given.

Fixed

Fix label code generation Remove all occurrences of a group in a MAPS when choosing the overwriting option. Fix resume when the training.tsv file is empty.

ClinicaDL 1.0.3 (release bugfix)

Fixed

• Fix import module for VAE architectures.

ClinicaDL 1.0.2

Changed

Core:

• Clinica version dependency.
• Update documentation, with proofreading.
• Change the test for extract.

Fixed

• Use the GPU option on the command line interface.
• Fix the train task when invoking the CLI.
• Replace the predict arguments passed through the CLI.
• Fix issue when using multiple cohorts.

ClinicaDL 1.0.1

Changed

Functionalities:

• extract improve this functionality with more expressive and easy to understand flags.
• Change the test for extract.

ClinicaDL 1.0

Welcome in ClinicaDL 1.0.0 !

This is a new version of ClinicaDL with some major changes in the source code. All the main pipelines have been refactored. ClinicaDL now used is now working with classes instead of function for easier maintenance and better scalability. We introduce in this version our new data structure called MAPS to unify ClinicaDL outputs.

This release also include a major command line refactoring with the introduction of Click library. Major changes in several pipelines interfaces have been implemented. For instance the preprocessing pipeline has been split. Train pipeline has also been refactored to reduce the number of options. The goal is to make ClinicaDL easier to use, to maintain, and adapt the command line to the MAPS.

ClinicaDL aim for more reproducibility: some configuration files are saved in the MAPS to reproduce experiments in the same condition (with same environment and same parameters). In addition, we added some options to fix the random processes seed and use Pytorch's latest enhancement for a deterministic behavior.

Other improvement and small fix have also been implemented.

Be careful this version breaks the backward compatibility with previous versions.

Added

Pipelines:

• New extract pipeline to convert nifti images in Pytorch tensors. This pipeline now saves a preprocessing json file with all the information needed for the train pipeline.

Core:

• New folder structure: MAPS (Model Analysis and Processing Structure) for ClinicaDL outputs.
• New class MapsManager for MAPS to ease the interface with the MAPS: launch various tasks such as training and prediction, save outputs, read files...
• Other new classes to make ClinicaDL code more scalable: SplitManager, TaskManager, Network...
• Now takes pet-linear images as possible data modality.
• Possibility to fix the seed for a deterministic behavior.

Other:

• New arguments for ROI in random search.

Changed

Pipelines:

• train pipeline has now a new command line. Please see the doc for more information.
• train now accept a TOML configuration file to simplify the command line.
• train resume is now named resume.
• random-search generate is now named random-search
• random-searchnow use TOML configuration file instead of JSON.
• generate pipeline to generate synthetic dataset (random, trivial and Shepp-Logan).
• quality-check pipeline to perform quality check on Clinica t1-linear and t1-volume outputs.

Core:

• Major refactoring of the source code repository organization: now ClinicaDL core respects the command line hierarchy as is in Clinica repository.
• Command line: we now use Click instead of Argparse.

Other:

• Console logs enhancement.

• Replace tensorboardx dependency with torch.utils.tensorboard to remove tensorboardx from requirement.

• GitHub repository name is now ClinicaDL and not AD-DL anymore.

• We will now use GitHub discussion instead of Google Group.

Deprecated

Removed

Pipelines:

• clinicadl preprocessing pipelines have been removed. They now have new names (see above).
• clinicadl preprocessing run has now completely been removed. Please use Clinica for preprocessing.
• train_from_json pipeline has been replaced with TOML configuration file.
• clinicadl random-search analysis pipeline has been removed.

Core:

• T1-extensive preprocessing is not supported by ClinicaDL anymore.

Fixed

• Fix multi-class classification.
• Fix classification when only one ROI is given.
• Fix tsvtools when diagnosis column value is not identical to tsv file name.
• Fix clinicadl tsvtool split when n_test is 1.
• Fix ROI when user provides 3D tensor for ROI.
• Fix the use of GPU on a machine with more than ong GPU (such a computer cluster). ClinicaDL will now select a free GPU that as permission.
• Fix an error where diagnosis column was automatically overwritten by the name of the file during split or kfold.

Security

ClinicaDL 0.2.2

Added

• New functionality clinicadl random-search analysis to obtain the histogram of the balanced accuracy over a random search folder.
• New functionality clinicadl train from_json to train a model with parameters defined in a JSON file.
• New functionality clinicadl train resume to resume a prematurely stopped training task.
• Possibility to learn the gray matter intensities with the binary classification during training, based on t1-volume outputs.
• Refactor code style using Black tool.

Changed

• Previous clinicadl random-search is now clinicadl random-search generate
• Cross-validation and computational arguments of clinicadl random-search generate are now defined in random_search.json.
• Remove tensorboardx dependency.

ClinicaDL 0.2.1

Added

• the multi_cohort flag in train allows to train on several CAPS at the same time.

Changed

• clinicadl train roi now allows any ROI defined by a mask.
• Update README.md to avoid duplicates.
• JSON files are added for clinicadl classify and clinicadl tsvtool getlabels|split|kfold

Removed

• Scripts and data related to MedIA publication.

ClinicaDL 0.2.0

Added

• New functionality clinicadl interpret to generate saliency maps linked to pretrained models based on groups of individual images.
• New functionality clinicadl random-search to sample random networks from a predefined hyperparameter space.
• Slice subparsers for autoencoder/cnn/multicnn to be homogeneous with other parsers.
• roi parser has now multicnn option.
• Add generic options to command line: --verbose, --version and --logname.

Changed

• Behaviour of clinicadl quality-check t1-volume.
• Simplify clinicadl tsvtools behaviour when using getlabels, split and analysis.
• Update documentation.

Fixed

• Fix broken file when running preprocessing in t1-extensive.