Search papers, labs, and topics across Lattice.
This paper benchmarks CNN-based SegResNet against hybrid transformer models (UNETR, SwinUNETR, UNETR++) for multi-organ segmentation on the RATIC dataset, a heterogeneous multi-institutional abdominal CT scan dataset. Surprisingly, SegResNet outperforms all transformer-based models in Dice Similarity Coefficient (DSC) across all organs, despite transformers' purported advantages in modeling long-range dependencies. UNETR++ was the best-performing transformer, while UNETR converged faster.
CNNs still reign supreme for medical image segmentation on heterogeneous datasets, beating out hybrid transformer models despite the latter's theoretical advantages.
Accurate multi-organ segmentation in abdominal CT scans is essential for computer-aided diagnosis and treatment. While convolutional neural networks (CNNs) have long been the standard approach in medical image segmentation, transformer-based architectures have recently gained attention due to their ability to model long-range dependencies. In this study, we systematically benchmark the three hybrid transformer-based models UNETR, SwinUNETR, and UNETR++ against a strong CNN baseline, SegResNet, for volumetric multi-organ segmentation on the heterogeneous RATIC dataset. The dataset comprises 206 annotated CT scans from 23 institutions worldwide, covering five abdominal organs. All models were trained and evaluated under identical preprocessing and training conditions using the Dice Similarity Coefficient (DSC) as the primary metric. The results show that the CNN-based SegResNet achieves the highest overall performance, outperforming all hybrid transformer-based models across all organs. Among the transformer-based approaches, UNETR++ delivers the most competitive results, while UNETR demonstrates notably faster convergence with fewer training iterations. These findings suggest that, for small- to medium-sized heterogeneous datasets, well-optimized CNN architectures remain highly competitive and may outperform hybrid transformer-based designs.
