Abstract
Recent advances in open-source magnetic resonance imaging (MRI) hardware, have significantly improved engineering of low field (LF) magnetic strengths, making LF viable for a broad range of clinical applications. It is regarded as a feasible imaging modality in low and middle-income countries (LMICs), where imaging resources remain highly constrained. However, LF MRI remains susceptible to noise, artifacts, coil variations, and acceleration-related degradation. These challenges raise critical questions as to whether the images acquired from these scanners can be trusted for safe clinical diagnosis and decision-making. To address this, a unified epistemic and aleatoric uncertainty enhanced autoencoder framework was used to analyse a scan acceptance rule using the open source OSI2 ONE k-space dataset from Leiden University Medical Center (LUMC). The result motivated the development of a two-axis uncertainty-based scan acceptance framework with separate percentile-derived thresholds. The framework successfully identified 42.3% of LUMC, unsafe low-rank reconstruction (LLR) scans while maintaining a zero false-rejection rate for safe scans.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MIRASOL_031.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=8PMkgwG97j
BibTex
@InProceedings{FunMbe_Towards_MICCAISAT2026,
author = { Funperefagha, Mberekpe Emmanuel AND Jimoh, Nafisa Opemi AND Tijjani, Usman AND Oyelami, Oyewole AND Okegbemi, Joshua Taiwo AND Usman, Hafsat Muhammed AND Obe, Olumide Olayinka AND Thisleton, William AND Raymond, Confidence AND Iorumbur, Aondona Moses AND Anazodo, Udunna C. AND Adewole, Maruf AND Adebayo, Adegboyega},
title = { { Towards the Development of a Clinical Safety Acceptance Rule for Low-Field MRI using Hybridized Uncertainty-Aware Autoencoders } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17265},
month = {pending},
page = {pending}
}