Frequency-Hierarchical Active k-Space Sampling for Diagnostic MRI

2026-07-22

AI Summary

This work addresses the challenge of balancing low-frequency (anatomical) and high-frequency (pathological detail) k-space sampling under limited acquisition budgets in accelerated MRI. The authors propose HieraSample, a novel framework that introduces a frequency hierarchy into active sampling by fully acquiring low-frequency data while dynamically selecting high-frequency samples via a Mamba-based policy network informed by dual classifiers encoding disease type and severity. The method employs cosine-annealed curriculum learning to progressively increase acceleration factors and incorporates a task-driven reinforcement learning reward based on reductions in classification cross-entropy. Evaluated on the fastMRI+ knee dataset, HieraSample achieves ACL diagnostic performance comparable to fully sampled references at 4×–10× acceleration, with up to a 20.4-percentage-point improvement in AUC for severity assessment.

Abstract

Active sampling for accelerated MRI must distribute a tight sampling budget across spatial frequencies that carry very different kinds of information. Low frequencies hold most of the anatomical context; high frequencies carry the fine details that drive pathology assessment. Existing active samplers either treat both regions identically or restrict the action space to entire Cartesian rows, which forces a poor compromise at high acceleration. We propose HieraSample, a task-driven framework built around this hierarchy. A cosine-annealed curriculum lowers the acceleration factor from 20x to 4x across 80 acquisition steps while keeping a fully-sampled low-frequency disk at every step; a Mamba-based policy then picks individual high-frequency coordinates from features extracted by dual disease and severity classifiers. The reward is the per-sample reduction in class-weighted cross-entropy after each action, so a positive reward corresponds directly to a more confident correct prediction. On the fastMRI+ knee benchmark, HieraSample matches the fully-sampled oracle on ACL diagnosis from 4x to 10x acceleration, and improves on a recent Cartesian baseline by as much as 20.4 AUC points on ACL severity.

Problem

• accelerated MRI
• active sampling
• spatial frequencies
• diagnostic imaging
• k-space sampling

Innovation

• active sampling
• frequency hierarchy
• Mamba-based policy
• task-driven MRI
• k-space curriculum learning