Reliable Evaluation of MRI Motion Correction: Dataset and Insights

Source

• Source name: OpenReview
• Source kind: paper
• Authors / organization: Kun Wang, Tobit Klug, Stefan Ruschke, Jan Kirschke, Reinhard Heckel
• Venue: ICLR.cc/2026/Conference
• Published: UNKNOWN
• Collected: 2026-06-28
• URL: https://openreview.net/forum?id=5PY8HR2Zz6
• PDF: https://openreview.net/pdf/fd893f4fdcf2cb5daeff39090bbff0521a23f390.pdf
• DOI: UNKNOWN

Plain-Language Brief

The paper discusses the difficulty in evaluating methods for correcting motion artifacts in medical imaging due to the lack of ground-truth data. To address this issue, they introduce PMoC3D, a dataset of real-world motion-corrupted 3D brain MRI data, and MoMRISim, a metric trained to evaluate motion reconstructions. They find that evaluating with real-world data combined with MoMRISim is the most reliable approach.

What the Document Says

• The paper discusses the importance of correcting motion artifacts in scientific and medical imaging as they impact image quality significantly.
• Evaluating deep learning-based and classical motion correction methods remains challenging due to the lack of accessible ground-truth target data.
• Three evaluation approaches are studied: real-world evaluation based on reference scans, simulated motion, and reference-free evaluation.
• To enable evaluation with real-world motion artifacts, they release PMoC3D, a dataset consisting of paired motion-corrupted 3D brain MRI data.
• They introduce MoMRISim, a feature-space metric trained for evaluating motion reconstructions.
• Each evaluation approach is assessed and real-world evaluation together with MoMRISim is found to be the most reliable.
• Evaluation based on simulated motion systematically exaggerates algorithm performance.
• Reference-free evaluation overrates oversmoothed deep learning outputs.

Key Concepts Mentioned

• medical imaging
• motion artifacts
• deep learning
• evaluation methods
• PMoC3D
• MoMRISim

Claims Made by the Document

• Real-world evaluation together with MoMRISim is the most reliable approach for evaluating motion correction methods.

Evidence Used

• They find real-world evaluation together with MoMRISim to be the most reliable based on their assessment of each evaluation approach.

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Correcting motion artifacts in scientific and medical imaging is important, as they significantly impact image quality.
However, evaluating deep learning-based and classical motion correction methods remains fundamentally difficult due to the lack of accessible ground-truth target data.
To address this challenge, we study three evaluation approaches: real-world evaluation based on reference scans, simulated motion, and reference-free evaluation, each with its merits and shortcomings.
To enable evaluation with real-world motion artifacts, we release PMoC3D, a dataset consisting of unprocessed $\textbf{P}$aired $\textbf{Mo}$tion-$\textbf{C}$orrupted $\textbf{3D}$ brain MRI data.
To advance evaluation quality, we introduce MoMRISim, a feature-space metric trained for evaluating motion reconstructions.
We assess each evaluation approach and find real-world evaluation together with MoMRISim, while not perfect, to be most reliable.
Evaluation based on simulated motion systematically exaggerates algorithm performance, and reference-free evaluation overrates oversmoothed deep learning outputs.