RF-HiT: Rectified Flow Hierarchical Transformer for General Medical Image Segmentation

TL;DR

RF-HiT introduces a computationally efficient hierarchical transformer for medical image segmentation, achieving high accuracy with minimal inference steps and low computational cost.

Contribution

It presents a novel rectified flow hierarchical transformer that combines efficiency with high segmentation performance, surpassing prior diffusion-based methods.

Findings

Achieves 91.27% mean Dice on ACDC dataset.

Requires only 10.14 GFLOPs and 13.6M parameters.

Operates with as few as three inference steps.

Abstract

Accurate medical image segmentation requires both long-range contextual reasoning and precise boundary delineation, a task where existing transformer- and diffusion-based paradigms are frequently bottlenecked by quadratic computational complexity and prohibitive inference latency. We propose RF-HiT, a Rectified Flow Hierarchical Transformer that integrates an hourglass transformer backbone with a multi-scale hierarchical encoder for anatomically guided feature conditioning. Unlike prior diffusion-based approaches, RF-HiT leverages rectified flow with efficient transformer blocks to achieve linear complexity while requiring only a few discretization steps. The model further fuses conditioning features across resolutions via learnable interpolation, enabling effective multi-scale representation with minimal computational overhead. As a result, RF-HiT achieves a strong efficiency-performance trade-off, requiring only 10.14 GFLOPs, 13.6M parameters, and inference in as few as three steps. Despite its compact design, RF-HiT attains 91.27% mean Dice on ACDC and 87.40% on BraTS 2021, achieving performance comparable to or exceeding that of significantly more intensive architectures. This demonstrates its strong potential as a robust, computationally efficient foundation for real-time clinical segmentation.