arXiv 2503.013062025-03-03eess.IVcs.AIcs.CVphysics.med-ph

从主张到证据：医学图像分割中CNN vs. Transformer vs. Mamba的统一框架与批判性分析

From Claims to Evidence: A Unified Framework and Critical Analysis of CNN vs. Transformer vs. Mamba in Medical Image Segmentation

Geneva University Hospital（日内瓦大学医院）；University of Groningen University Medical Center Groningen（格罗宁根大学医学中心）；University of Southern Denmark（南丹麦大学）

Pooya Mohammadi Kazaj, Giovanni Baj, Yazdan Salimi, Anselm W. Stark, Waldo Valenzuela, George CM. Siontis, Habib Zaidi, Mauricio Reyes, Christoph Graeni, Isaac Shiri

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

提出基于nnUNet的开源基准框架nnUZoo，结合CNN、Transformer和Mamba架构（含新提出的X2Net），在六类医学数据集上公平比较发现CNN兼顾速度与精度，Mamba架构参数更少且精度相当但训练时间更长。

摘要：

尽管已有许多用于医学图像分割的架构被提出，但使用最先进的模型网络（如nnUNet）实现具有竞争力的性能，仍有进一步创新的空间。在本研究中，我们引入了nnUZoo，这是一个基于nnUNet构建的开源基准测试框架，它整合了多种深度学习架构，包括CNN、Transformer和基于Mamba的模型。利用该框架，我们提供了一个公平的比较，以揭示不同医学图像分割任务中的性能主张。此外，为了丰富基准测试，我们探索了五种基于Mamba和Transformer的新架构，统称为X2Net，并将它们集成到nnUZoo中以进行进一步评估。所提出的模型结合了传统U2Net、nnUNet、CNN、Transformer和Mamba层与架构的特征，称为X2Net（包括UNETR2Net (UNETR)、SwT2Net (SwinTransformer)、SS2D2Net (SwinUMamba)、Alt1DM2Net (LightUMamba)和MambaND2Net (MambaND)）。我们在六个多样化的医学图像分割数据集上广泛评估了不同模型的性能，包括显微镜、超声、CT、MRI和PET，覆盖了各种身体部位、器官和标签。我们以Dice分数和计算效率为指标，将它们的性能与基线模型U2Net和nnUNet进行了比较。像nnUNet和U2Net这样的CNN模型展示了速度和准确性，使其成为医学图像分割任务的有效选择。基于Transformer的模型虽然在某些成像模态上很有前景，但表现出了很高的计算成本。所提出的基于Mamba的X2Net架构（SS2D2Net）实现了具有竞争力的准确率，与nnUNet和U2Net没有显著差异，同时使用了更少的参数。然而，它们需要显著更长的训练时间，凸显了模型效率与计算成本之间的权衡。

英文摘要：

While numerous architectures for medical image segmentation have been proposed, achieving competitive performance with state-of-the-art models networks such as nnUNet, still leave room for further innovation. In this work, we introduce nnUZoo, an open source benchmarking framework built upon nnUNet, which incorporates various deep learning architectures, including CNNs, Transformers, and Mamba-based models. Using this framework, we provide a fair comparison to demystify performance claims across different medical image segmentation tasks. Additionally, in an effort to enrich the benchmarking, we explored five new architectures based on Mamba and Transformers, collectively named X2Net, and integrated them into nnUZoo for further evaluation. The proposed models combine the features of conventional U2Net, nnUNet, CNN, Transformer, and Mamba layers and architectures, called X2Net (UNETR2Net (UNETR), SwT2Net (SwinTransformer), SS2D2Net (SwinUMamba), Alt1DM2Net (LightUMamba), and MambaND2Net (MambaND)). We extensively evaluate the performance of different models on six diverse medical image segmentation datasets, including microscopy, ultrasound, CT, MRI, and PET, covering various body parts, organs, and labels. We compare their performance, in terms of dice score and computational efficiency, against their baseline models, U2Net, and nnUNet. CNN models like nnUNet and U2Net demonstrated both speed and accuracy, making them effective choices for medical image segmentation tasks. Transformer-based models, while promising for certain imaging modalities, exhibited high computational costs. Proposed Mamba-based X2Net architecture (SS2D2Net) achieved competitive accuracy with no significantly difference from nnUNet and U2Net, while using fewer parameters. However, they required significantly longer training time, highlighting a trade-off between model efficiency and computational cost.

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。