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

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.

U-Net-based Models for Skin Lesion Segmentation: More Attention and Augmentation

According to WHO[1], since the 1970s, diagnosis of melanoma skin cancer has been more frequent. However, if detected early, the 5-year survival rate for melanoma can increase to 99 percent. In this regard, skin lesion segmentation can be pivotal in monitoring and treatment planning. In this work, ten models and four augmentation configurations are trained on the ISIC 2016 dataset. The performance and overfitting are compared utilizing five metrics. Our results show that the U-Net-Resnet50 and the R2U-Net have the highest metrics value, along with two data augmentation scenarios. We also investigate CBAM and AG blocks in the U-Net architecture, which enhances segmentation performance at a meager computational cost. In addition, we propose using pyramid, AG, and CBAM blocks in a sequence, which significantly surpasses the results of using the two individually. Finally, our experiments show that models that have exploited attention modules successfully overcome common skin lesion segmentation problems. Lastly, in the spirit of reproducible research, we implement models and codes publicly available.