README.md

---
title: clinicadl
emoji: 🔥
colorFrom: purple
colorTo: purple
sdk: static
pinned: true
license: mit
---

# What is ClinicaDL ?

ClinicaDL is an open-source deep learning software for reproducible neuroimaging processing. It can be seen as the deep learning extension of 
[Clinica](https://aramislab.paris.inria.fr/clinica/docs/public/latest/CAPS/Introduction/), an open-source Python library for neuroimaging preprocessing 
and analysis. The combination of ClinicaDL and Clinica allows performing an end-to-end neuroimaging analysis, from the download of raw data sets to the 
interpretation of trained networks, including neuroimaging preprocessing, quality check, label definition, architecture search, and network training and 
evaluation.
ClinicaDL has been implemented to bring answers to three common issues encountered by deep learning users who are not always familiar with neuroimaging data: 
- accessing properly formatted and pre-processed datasets can be difficult, which can be partly tackled by a dataset format established by the community: the Brain Imaging
  Data Structure (BIDS) - methodological flaws in many studies which results are contaminated by data leakage,
- a lack of reproducibility that discredits results,
Employing ClinicaDL serves as an initial measure to avoid such prevalent problems.

This library was at first developed from the AD-DL project, a GitHub repository hosting the source code of a scientific publication on the deep learning classification 
of brain images in the context of Alzheimer's disease. This is why some functions of ClinicaDL can still be specific to Alzheimer's disease context.
For moreinformation on this clinical context, please refer to our [tutorial](https://aramislab.paris.inria.fr/clinicadl/tuto/2023/html/index.html).
If you are new to ClinicaDL, please consider reading the [First steps section](https://clinicadl.readthedocs.io/en/latest/Introduction/) before starting your project!