DRIFT: Difficulty-aware Rectified Flows for Through-plane MRI Super-Resolution

2026-07-18

AI Summary

This work addresses the challenges of stair-step artifacts in anisotropic MRI caused by large slice thickness and the inherent trade-off between efficiency and fidelity in super-resolution reconstruction. To this end, the authors propose a two-stage, slice-thickness-conditioned rectified flow framework. In the first stage, an anatomical projection network generates a deterministic high-resolution initial estimate. The second stage introduces a physics-aware difficulty metric to adaptively schedule the number of ODE integration steps and incorporates a consistency-based endpoint trajectory alignment loss to ensure reconstruction stability across varying slice thicknesses. The proposed method achieves superior reconstruction quality while maintaining low inference cost, effectively balancing computational efficiency and image fidelity.

Abstract

Magnetic Resonance Imaging (MRI) is often acquired with anisotropic resolution to reduce scan time, producing stair-step artifacts along the through-plane direction. In through-plane MRI super-resolution, an efficiency-fidelity trade-off arises: feed-forward regressors are fast but oversmooth at large slice-thicknesses, while sampling-based methods improve fidelity at high inference cost. We propose DRIFT, a two-stage thickness-conditioned rectified flow framework for through-plane MRI super-resolution with continuous input slice-thickness. Stage 1 employs an Anatomical Projection Network (APN) to map low-resolution patches to a coarse high-resolution manifold, providing a deterministic anatomical initialization that shortens the residual transport of Stage 2 and stabilizes slice-wise refinement. Stage 2 refines details via rectified flow and introduces a Physics-Aware Difficulty (PAD) metric derived from slice-thickness induced through-plane bandwidth deficit to guide an Adaptive Integration Scheduler (AIS), allocating ODE steps by thickness. A Consistent Endpoint Trajectory Alignment (CETA) loss enforces thickness-consistent reconstructions. Experiments show that DRIFT outperforms super-resolution baselines while reducing inference cost. Code, models, and interactive demos are available at https://yoonseokchoi-ai.github.io/drift-eccv2026/.

Problem

• through-plane MRI super-resolution
• anisotropic resolution
• stair-step artifacts
• efficiency-fidelity trade-off
• slice-thickness

Innovation

• rectified flow
• through-plane super-resolution
• thickness-conditioned
• adaptive integration scheduler
• physics-aware difficulty