arXiv 2507.161222025-07-25eess.IVcs.CV

MLRU++：用于高效3D医学图像分割的带注意力的多尺度轻量级残差UNETR++

MLRU++: Multiscale Lightweight Residual UNETR++ with Attention for Efficient 3D Medical Image Segmentation

AI Research Lab, Department of Computer Science（人工智能研究实验室、计算机科学系）；Biomedical & Translational Sciences, Sanford School of Medicine（生物医学与转化科学、桑福德医学院）；University Of South Dakota（南达科他大学）

Nand Kumar Yadav, Rodrigue Rizk, William CW Chen, KC Santosh

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

针对3D医学图像分割中精度与计算成本难以兼顾的问题，提出轻量级混合架构MLRU++，通过LCBAM和M2B提升多尺度特征表达，在多个基准上以更少参数达到最先进性能。

摘要：

准确且高效的医学图像分割至关重要，但由于解剖结构的变异性以及体数据的高计算需求而具有挑战性。近期的CNN-Transformer混合架构取得了最先进的结果，却显著增加了复杂度。本文提出MLRU++，这是一种多尺度轻量级残差UNETR++架构，旨在平衡分割精度与计算效率。它引入了两项关键创新：轻量级通道与瓶颈注意力模块（LCBAM），以极小开销增强上下文特征编码；以及解码器中的多尺度瓶颈块（M2B），通过多分辨率特征聚合捕获细粒度细节。在四个公开基准数据集（Synapse、BTCV、ACDC和Decathlon Lung）上的实验表明，MLRU++取得了最先进性能，平均Dice分数分别为87.57%（Synapse）、93.00%（ACDC）和81.12%（Lung）。与现有领先模型相比，MLRU++在Synapse和ACDC上将Dice分数分别提高5.38%和2.12%，同时显著减少参数量和计算成本。评估LCBAM和M2B的消融研究进一步证实了所提架构组件的有效性。结果表明，MLRU++为3D医学图像分割任务提供了实用且高性能的解决方案。源代码可在https://github.com/1027865/MLRUPP获取。

英文摘要：

Accurate and efficient medical image segmentation is crucial but challenging due to anatomical variability and high computational demands on volumetric data. Recent hybrid CNN-Transformer architectures achieve state-of-the-art results but add significant complexity. In this paper, we propose MLRU++, a Multiscale Lightweight Residual UNETR++ architecture designed to balance segmentation accuracy and computational efficiency. It introduces two key innovations: a Lightweight Channel and Bottleneck Attention Module (LCBAM) that enhances contextual feature encoding with minimal overhead, and a Multiscale Bottleneck Block (M2B) in the decoder that captures fine-grained details via multi-resolution feature aggregation. Experiments on four publicly available benchmark datasets (Synapse, BTCV, ACDC, and Decathlon Lung) demonstrate that MLRU++ achieves state-of-the-art performance, with average Dice scores of 87.57% (Synapse), 93.00% (ACDC), and 81.12% (Lung). Compared to existing leading models, MLRU++ improves Dice scores by 5.38% and 2.12% on Synapse and ACDC, respectively, while significantly reducing parameter count and computational cost. Ablation studies evaluating LCBAM and M2B further confirm the effectiveness of the proposed architectural components. Results suggest that MLRU++ offers a practical and high-performing solution for 3D medical image segmentation tasks. Source code is available at: https://github.com/1027865/MLRUPP

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。