Untrained Convolutional Neural Networks as Feature Extractors for Structural MRI
Arel Encin, Inés Gonzalez Pepe, Yohan Chatelain, Erin W. Dickie, Tristan Glatard
The paper shows that untrained convolutional neural networks can extract features from structural MRI data that perform as well as or better than pretrained models in three tasks across three datasets.
bioRxiv · Jun 10, 2026 · PreprintNeuroscience