Abstract
Portable ultra-low-field (ULF) MRI is extending pediatric neuroimaging to low-resource settings where access to conventional scanners is scarce, yet its intrinsically weak signal makes it highly vulnerable to artifacts. Relia-ble quality assessment (QA), deciding which artifacts are present and how severe they are, is therefore an imperative for any downstream use. We ad-dress this challenge beginning with the observation that the artifacts most difficult to detect—Zipper, Motion, and Banding—arise in the frequency domain (k-space) of the MRI signal; hence, using this domain will perform better than approximating them as overlays on reconstructed images. Guid-ed by this principle, we develop a QA system built from three coordinated components, comprising (i) physics-based artifact synthesis performed di-rectly in 2-D k-space, which mimics the intrinsic frequency signatures of the acquisition; (ii) an orientation-aware classifier, since the scan plane is strongly predictive of which artifacts occur; and (iii) an attention-based ag-gregation that condenses slice-level predictions by concentrating on the few slices that exhibit an artifact. A fine-tuned MambaOut model achieves 0.833 micro-F1 and 0.564 macro-F1 on the challenge’s validation split. Our contribution lies in integrating physics-aware synthesis and orientation conditioning, tailored to the demands of ultra-low-field pediatric QA.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/LISA_019.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=WjnY1ee9S1
BibTex
@InProceedings{HwaEun_PhysicsAware_MICCAISAT2026,
author = { Hwang, Eunchan},
title = { { Physics-Aware k-Space Artifact Simulation for Quality Assessment of Ultra-Low-Field Pediatric Brain MRI } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17257},
month = {pending},
page = {pending}
}