Transmittance Map to Fiber Orientation Map with Generative Models

Overview

This project explores the use of deep learning generative models to process 3D-Polarized Light Imaging (3D-PLI) data of vervet monkey brains. The goal is twofold:

1. FOM Generation: Generate Fiber Orientation Map (FOM) patches using a generative model.
2. Transmittance-to-FOM Translation: Implement and evaluate an image-to-image translation model that generates FOM maps from corresponding transmittance maps.

This work involves preprocessing large-scale imaging datasets, training generative models (GANs, Pix2Pix, etc.), and evaluating model outputs using both qualitative visualizations and quantitative metrics.

Getting started

Poetry

if poetry not installed download it

• install guide
• poetry install to install al packages
• poetry shell start virtual environment
• poetry add [package] add new package

Dataset

The original dataset is publicly available via the EBRAINS platform of the Human Brain Project (Axer et al., 2020; https://doi.org/10.25493/AFR3-KDK).

Reference:

Axer, M., Gräßel, D., Palomero-Gallagher, N., Takemura, H., Jorgensen, M. J., Woods, R., & Amunts, K. (2020). Images of the nerve fiber architecture at micrometer-resolution in the vervet monkey visual system [Data set]. EBRAINS. https://doi.org/10.25493/AFR3-KDK