Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BRIQA: Balanced Reweighting in Image Quality Assessment of Pediatric Brain MRI

Alya Almsouti, Ainur Khamitova, Darya Taratynova, Mohammad Yaqub

TL;DR

The paper tackles automated artifact-severity assessment in pediatric brain MRI acquired on low-field systems, where class imbalance across seven artifact types complicates learning. It introduces BRIQA, a multitask framework that uses gradient-based loss reweighting, axis prediction, and a rotating batching strategy to balance exposure to underrepresented severity levels, with an emphasis on architectural diversity across artifact types. Empirical results show that no single backbone is best for all artifacts, but rotating batching combined with cross-entropy loss yields strong macro-averaged performance, improving mean macro F1 from $0.659$ to $0.706$ and delivering notable gains on Noise, Zipper, Positioning, and Contrast artifacts; frequency-domain fusion offers complementary improvements. The approach has practical implications for robust automated IQA in heterogeneous, low-resource MRI environments, though it requires larger multi-center validation due to dataset limitations and the overhead of maintaining multiple models.

Abstract

Assessing the severity of artifacts in pediatric brain Magnetic Resonance Imaging (MRI) is critical for diagnostic accuracy, especially in low-field systems where the signal-to-noise ratio is reduced. Manual quality assessment is time-consuming and subjective, motivating the need for robust automated solutions. In this work, we propose BRIQA (Balanced Reweighting in Image Quality Assessment), which addresses class imbalance in artifact severity levels. BRIQA uses gradient-based loss reweighting to dynamically adjust per-class contributions and employs a rotating batching scheme to ensure consistent exposure to underrepresented classes. Through experiments, no single architecture performs best across all artifact types, emphasizing the importance of architectural diversity. The rotating batching configuration improves performance across metrics by promoting balanced learning when combined with cross-entropy loss. BRIQA improves average macro F1 score from 0.659 to 0.706, with notable gains in Noise (0.430), Zipper (0.098), Positioning (0.097), Contrast (0.217), Motion (0.022), and Banding (0.012) artifact severity classification. The code is available at https://github.com/BioMedIA-MBZUAI/BRIQA.

BRIQA: Balanced Reweighting in Image Quality Assessment of Pediatric Brain MRI

TL;DR

The paper tackles automated artifact-severity assessment in pediatric brain MRI acquired on low-field systems, where class imbalance across seven artifact types complicates learning. It introduces BRIQA, a multitask framework that uses gradient-based loss reweighting, axis prediction, and a rotating batching strategy to balance exposure to underrepresented severity levels, with an emphasis on architectural diversity across artifact types. Empirical results show that no single backbone is best for all artifacts, but rotating batching combined with cross-entropy loss yields strong macro-averaged performance, improving mean macro F1 from to and delivering notable gains on Noise, Zipper, Positioning, and Contrast artifacts; frequency-domain fusion offers complementary improvements. The approach has practical implications for robust automated IQA in heterogeneous, low-resource MRI environments, though it requires larger multi-center validation due to dataset limitations and the overhead of maintaining multiple models.

Abstract

Assessing the severity of artifacts in pediatric brain Magnetic Resonance Imaging (MRI) is critical for diagnostic accuracy, especially in low-field systems where the signal-to-noise ratio is reduced. Manual quality assessment is time-consuming and subjective, motivating the need for robust automated solutions. In this work, we propose BRIQA (Balanced Reweighting in Image Quality Assessment), which addresses class imbalance in artifact severity levels. BRIQA uses gradient-based loss reweighting to dynamically adjust per-class contributions and employs a rotating batching scheme to ensure consistent exposure to underrepresented classes. Through experiments, no single architecture performs best across all artifact types, emphasizing the importance of architectural diversity. The rotating batching configuration improves performance across metrics by promoting balanced learning when combined with cross-entropy loss. BRIQA improves average macro F1 score from 0.659 to 0.706, with notable gains in Noise (0.430), Zipper (0.098), Positioning (0.097), Contrast (0.217), Motion (0.022), and Banding (0.012) artifact severity classification. The code is available at https://github.com/BioMedIA-MBZUAI/BRIQA.

Paper Structure

This paper contains 8 sections, 4 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Dataset
  3. Model Description
  4. Experimental Setup
  5. Results
  6. Discussion
  7. Conclusion
  8. Disclosure of Interests.

Figures (2)

  • Figure 1: Scans from multiple patients obtained with the 0.064T Hyperfine SWOOP system, showing severe artifacts in different anatomical planes.
  • Figure 2: Rotating batch configuration. Samples from class 0 are drawn using a rotating buffer mechanism, where their indices are cyclically shifted across epochs according to a modular offset. Note that for each sample, we take the remainder of the calculated index with respect to $N_0$ to achieve the cycling effect; this operation is not shown in the figure.