Data and Code from "MR-AIV reveals in-vivo brain-wide fluid flow with physics-informed AI"

This repository contains the original DCE-MRI data (nifti_original_data) used in training the neural network model described in "Mapping and quantifying deep brain fluid flow in vivo with AI", along with the final inferred speeds (nifti_results).  Data.zip includes 1.) the synthetic data used to validate the model, and 2.) the filtered real DCE-MRI data, with the initial guesses of the velocity and permeability.

There are three different synthetic data sets ("smooth", "sharp", and "realistic"), corresponding to the permeability map used to generate the data, and five real data sets corresponding to different mice, referred to as either M* or mouse01. 

The code required to train and visualize the results is contained in the "MR-AIV_Code.zip" file. Once unzipped, please follow the instructions detailed in the README.md file.

The algorithm used to obtain the initial guess of the velocity via front tracking is FrontTracker3D.m.