首页 > AI前沿 > A thorough review on Deep Learning approaches to 2D-to-3D Registration in Medical Imaging

A thorough review on Deep Learning approaches to 2D-to-3D Registration in Medical Imaging

arXiv机器学习 2026-11-13 03:52 5 阅读 查看原文
用户现在需要把给定的新闻正文排版成语义化的HTML片段,首先看内容是学术类的综述内容,首先第一个自然段整体是一个段落吗?不对,先捋清楚结构:首先第一段是讲2D-to-3D图像配准的应用场景,传统方法的不足,深度学习的优势,然后引出这篇综述的内容。然后接下来的内容,首先有没有需要分小节的?不对,先看原文的段落,哦原文是一整段?不对,不对,再仔细看原文,哦原文是连续的内容?不对,不对,先看要求,每个自然段落用p包裹,长文用h2/h3划分小节。首先,先梳理结构:首先,开头的背景部分可以作为第一段?不对,等下原文的内容是:首先开头是“2D-to-3D image registration...real-time performance.”然后接下来“This review provides...real-time performance?不对,哦原文的内容是: 首先第一部分是背景:2D-to-3D image registration is a key component...potential for real-time performance. 然后接下来是综述的内容介绍:This review provides a comprehensive analysis...across studies. 然后是研究发现:Across the 41 included studies...X-ray-to-CT registration. 然后是结论和展望:We conclude by identifying...for representation learning. 哦不对,等下有没有小节划分的必要?比如可以分几个小节?比如第一个h2可以是“研究背景与综述概述”?不对,不对,要求是保留原文所有信息,不得增删改内容!哦对,不能加原文没有的内容,所以小节标题不能自己编?哦对哦,规则4说保留原文所有信息,不得增删改内容,那如果原文没有小标题,那是不是不用加h2/h3?不对,再看规则2:长文用

/

划分小节,但是如果原文没有现成的小节标题,那不能自己加啊,因为不能增内容。哦对哦,那原文是连续的一段?不对,再看原文的内容,是不是自然的段落?哦原文是一整段吗?不,等下原文的内容是: 整段是:2D-to-3D image registration is a key component for several medical imaging purposes such as intraoperative guidance, enabling the alignment between preoperative 3D volumes and real-time 2D images. While conventional optimisation-based methods often struggle with non-convex objectives, sensitivity to initialisation, and high computational demands, deep learning (DL) has emerged as a promising alternative, offering improved alignment robustness and the potential for real-time performance. This review provides a comprehensive analysis of DL frameworks developed to address the challenges inherent in 2D-to-3D registration. We first outline how core concepts of DL are adapted to the registration problem. We then examine recent methodological advancements from multiple perspectives, including pipeline design, architectural innovations, and domain adaptation strategies. Particular attention is given to quantitative metrics, highlighting how performance is defined and measured across studies. Across the 41 included studies, we find that direct regression is the most common pipeline (23 studies, against 18 iterative ones), that hybrid CNN-attention architectures are gaining ground, and that research remains heavily concentrated on X-ray-to-CT registration. We conclude by identifying the absence of per-case uncertainty estimation and the strong neglect of temporal or sequential modelling as the two methodological gaps most consequential for clinical readiness, and discuss state-space models as a way to add temporal memory within intraoperative latency budgets. Further opportunities lie in more principled domain adaptation and in the integration of foundation models for representation learning. 哦,这是连续的一段?不对,不对,有没有可以分的?哦不对,规则说不能增删改内容,所以不能自己加小标题,对吧?那是不是所有内容放在

里?不对,等下,有没有重要的句子可以加?比如重要的结论?比如“We conclude by identifying the absence of per-case uncertainty estimation and the strong neglect of temporal or sequential modelling as the two methodological gaps most consequential for clinical readiness, and discuss state-space models as a way to add temporal memory within intraoperative latency budgets.”还有“deep learning (DL) has emerged as a promising alternative, offering improved alignment robustness and the potential for real-time performance.”这些可以加strong?不对,或者看有没有列表?原文没有列表,引用也没有,代码也没有。那结构就是: 首先,有没有小节?哦不对,原文没有小标题,所以不用加h2/h3,对吧?那整个正文是一个自然段落?不对,等下再仔细看,是不是可以分成几个p?比如第一个p是讲背景和传统方法与DL的对比,第二个p是讲综述的内容安排,第三个p是研究发现,第四个p是结论和展望?不对,但是原文是连续的,是不是原来的自然段落就是一整段?哦,用户给的正文是连续的一段吗?看用户给的正文: 正文如下: 2D-to-3D image registration is a key component for several medical imaging purposes such as intraoperative guidance, enabling the alignment between preoperative 3D volumes and real-time 2D images. While conventional optimisation-based methods often struggle with non-convex objectives, sensitivity to initialisation, and high computational demands, deep learning (DL) has emerged as a promising alternative, offering improved alignment robustness and the potential for real-time performance. This review provides a comprehensive analysis of DL frameworks developed to address the challenges inherent in 2D-to-3D registration. We first outline how core concepts of DL are adapted to the registration problem. We then examine recent methodological advancements from multiple perspectives, including pipeline design, architectural innovations, and domain adaptation strategies. Particular attention is given to quantitative metrics, highlighting how performance is defined and measured across studies. Across the 41 included studies, we find that direct regression is the most common pipeline (23 studies, against 18 iterative ones), that hybrid CNN-attention architectures are gaining ground, and that research remains heavily concentrated on X-ray-to-CT registration. We conclude by identifying the absence of per-case uncertainty estimation and the strong neglect of temporal or sequential modelling as the two methodological gaps most consequential for clinical readiness, and discuss state-space models as a way to add temporal memory within intraoperative latency budgets. Further opportunities lie in more principled domain adaptation and in the integration of foundation models for representation learning. 哦,确实是一整段,但是规则2说每个自然段落用p包裹,那如果原文是一整段,就用一个p?不对,但是可以加strong对吧?比如重要的句子,比如DL的优势,还有研究发现的核心内容,还有结论的核心点。等下,规则3说重要句子可用strong,那可以给重要的句子加。比如:

2D-to-3D image registration is a key component for several medical imaging purposes such as intraoperative guidance, enabling the alignment between preoperative 3D volumes and real-time 2D images. While conventional optimisation-based methods often struggle with non-convex objectives, sensitivity to initialisation, and high computational demands, deep learning (DL) has emerged as a promising alternative, offering improved alignment robustness and the potential for real-time performance. This review provides a comprehensive analysis of DL frameworks developed to address the challenges inherent in 2D-to-3D registration. We first outline how core concepts of DL are adapted to the registration problem. We then examine recent methodological advancements from multiple perspectives, including pipeline design, architectural innovations, and domain adaptation strategies. Particular attention is given to quantitative metrics, highlighting how performance is defined and measured across studies. Across the 41 included studies, we find that direct regression is the most common pipeline (23 studies, against 18 iterative ones), that hybrid CNN-attention architectures are gaining ground, and that research remains heavily concentrated on X-ray-to-CT registration. We conclude by identifying the absence of per-case uncertainty estimation and the strong neglect of temporal or sequential modelling as the two methodological gaps most consequential for clinical readiness, and discuss state-space models as a way to add temporal memory within intraoperative latency budgets. Further opportunities lie in more principled domain adaptation and in the integration of foundation models for representation learning.

不对,但是有没有需要分h2/h3的?哦,原文没有小标题,所以不能自己加,因为不能增内容,对吧?对,规则4说不得增删改内容,所以不能自己编小标题。那是不是这样就可以?等下,再检查有没有其他结构,比如有没有列表?原文没有,引用没有,代码没有,所以不需要ul或者blockquote、pre。那就是一个p,里面给重要的句子加strong。不对,等下,是不是可以分成几个p?比如按内容逻辑分: 第一个p:讲配准的应用场景,传统方法的问题,DL的优势。 第二个p:讲这篇综述的内容安排:介绍DL概念适配,方法进展,定量指标。 第三个p:讲研究发现。 第四个p:讲结论和展望。 这样是不是符合自然段落?哦,用户给的原文是一整段,但是是不是可以按语义分成多个p?因为规则2说每个自然段落用p