面向可变形医学图像配准中组合爆炸问题的动态流网络
Dynamic Stream Network for Combinatorial Explosion Problem in Deformable Medical Image Registration
把「双输入特征组合爆炸」显式建模为问题，并用动态感受野（AdSB）与动态权重注意力（DySA）筛选相关特征组合，这一思路可迁移到光流、立体匹配、多帧融合等任何双/多输入密集预测任务的特征交互设计上。
原文依据 the combination relationships between features grow exponentially, ultimately the model considers more irrelevant features during the feature modeling process
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论双输入导致的组合爆炸问题是可变形医学图像配准（DMIR）中的关键挑战。由于 DMIR 同时处理两幅图像作为输入，特征之间的组合关系呈指数增长，最终模型在特征建模过程中会考虑更多无关特征。在网络感受野与权重中引入动态性，可使模型剔除无关的特征组合并建模潜在的特征组合关系。本文提出动态流网络（DySNet），使感受野与权重能够被动态调整，从而最终使模型忽略无关特征组合并建模潜在的特征关系。本文包含两项关键创新：1）自适应流域（AdSB）模块动态调整感受野的形状，从而使模型聚焦于高度相关的特征关系；2）动态流注意力（DySA）机制生成动态权重，以搜索更有价值的特征关系。大量实验表明，DySNet 持续优于最先进的 DMIR 方法，凸显了其出色的泛化能力。代码已发布于网站：https://github.com/ShaochenBi/DySNet。
英文原文摘要 1363 字符
Combinatorial explosion problem caused by dual inputs presents a critical challenge in Deformable Medical Image Registration (DMIR). Since DMIR processes two images simultaneously as input, the combination relationships between features grow exponentially, ultimately the model considers more irrelevant features during the feature modeling process. Introducing dynamics in the receptive fields and weights of the network enables the model to eliminate the irrelevant features combination and model the potential feature combination relationships. In this paper, we propose the Dynamic Stream Network (DySNet), which enables the receptive fields and weights to be dynamically adjusted. This ultimately enables the model to ignore irrelevant feature combinations and model the potential feature relationships. With two key innovations: 1) Adaptive Stream Basin (AdSB) module dynamically adjusts the shape of the receptive field, thereby enabling the model to focus on the highly correlated feature relationships. 2) Dynamic Stream Attention (DySA) mechanism generates dynamic weights to search for more valuable feature relationships. Extensive experiments have shown that DySNet consistently outperforms the most advanced DMIR methods, highlighting its outstanding generalization ability. Our code is released on the website: https://github.com/ShaochenBi/DySNet.