Can We Encode Intra- and Inter-Variability with Log Jacobian Maps Derived from Brain Morphological Deformations Using Pediatric MRI Scans?

Example of log Jacobian maps used for training for both intra and inter pairs

Environment Details

conda env create -f environment.yml
conda activate tf2

Whole Pipeline

The pipeline involves rigid registration of intra and inter pairs from 96 subjects, followed by ANTs SyN registration for elastic deformation extraction. The deformation is transformed into Jacobian maps and log-scaled. Using 10-fold cross-validation, 10 sets with no overlaps were generated, resulting in 598 and 538 pairs for the ia_r and ias_r experiments respectively. In the overlap scenario, the same number of pairs were randomly selected whilst also ensuring a balanced 50/50 split between intra and inter pairs. These pairs are then fed into a 17-layer CNN. It is trained and tested with either overlap or no overlap in a 10-fold cross-validation scheme

Log Jacobian Analysis

Non-distinct distributions of the average absolute log Jacobian value amongst the selected pairs for both no overlap (left) and overlap (right) scenarios with regards to the age interval per pair colored by their pair type (or label). The ias_r experiment (538 pairs) results are shown on top and ia_r (598 pairs) on the bottom

Citing

MIDL 2024 Proceedings

@inproceedings{
dimitrijevic2024can,
title={Can We Encode Intra- and Inter-Variability with Log Jacobian Maps Derived from Brain Morphological Deformations Using Pediatric {MRI} Scans?},
author={Andjela Dimitrijevic and Fanny D{\'e}geilh and Benjamin De Leener},
booktitle={Medical Imaging with Deep Learning},
year={2024},
url={https://openreview.net/forum?id=4rKfNtCwmh}
}