DiffBench#

DiffBench is a Python library for benchmarking machine learning pipelines for diffusion MRI.

It provides a reproducible and modular framework for preparing diffusion MRI data, extracting features, training predictive models, evaluating their performance, and comparing experiments.

Benchmark workflow#

DiffBench follows a modular workflow from diffusion MRI data to benchmark results:

Diffusion MRI data
        │
        ▼
1. Preprocessing
        │
        ▼
2. Data Preparation
        │
        ▼
3. Feature Extraction
        │
        ▼
4. Model Training & Evaluation
        │
        ▼
5. Analysis

Each stage can be configured and used independently depending on the experiment.

1. Preprocessing

Preprocess diffusion MRI data and generate the inputs required by the benchmark.

Preprocessing

2. Data Preparation

Prepare diffusion MRI representations required by the benchmark.

Data Preparation

3. Feature Extraction

Transform prepared imaging data into model-ready features.

Feature Extraction

4. Model Training

Train and compare machine learning and deep learning models.