Shows that deep learning models for medical imaging (MRI) can be trained using synthetic quaternion Julia fractals instead of sensitive human clinical data.
April 1, 2026
Original Paper
Training deep learning based dynamic MR image reconstruction using synthetic fractals
arXiv · 2603.29922
The Takeaway
Bypasses the extreme privacy and licensing hurdles of medical AI. The ability to achieve comparable clinical reconstruction performance using purely synthetic, non-human-derived geometric data is a major breakthrough for data-starved medical domains.
From the abstract
Purpose: To investigate whether synthetically generated fractal data can be used to train deep learning (DL) models for dynamic MRI reconstruction, thereby avoiding the privacy, licensing, and availability limitations associated with cardiac MR training datasets. Methods: A training dataset was generated using quaternion Julia fractals to produce 2D+time images. Multi-coil MRI acquisition was simulated to generate paired fully sampled and radially undersampled k-space data. A 3D UNet deep artefa