Cats: Complementary CNN and Transformer Encoders for Segmentation

TL;DR

This paper introduces a hybrid 3D biomedical image segmentation model combining CNN and transformer encoders, achieving superior accuracy by leveraging both local and global feature extraction.

Contribution

The paper presents a novel dual-encoder architecture that fuses CNN and transformer features for improved 3D medical image segmentation performance.

Findings

Outperforms state-of-the-art models on three public datasets.

Achieves higher Dice scores across multiple challenging segmentation tasks.

Effectively combines local and global information for better segmentation accuracy.

Abstract

Recently, deep learning methods have achieved state-of-the-art performance in many medical image segmentation tasks. Many of these are based on convolutional neural networks (CNNs). For such methods, the encoder is the key part for global and local information extraction from input images; the extracted features are then passed to the decoder for predicting the segmentations. In contrast, several recent works show a superior performance with the use of transformers, which can better model long-range spatial dependencies and capture low-level details. However, transformer as sole encoder underperforms for some tasks where it cannot efficiently replace the convolution based encoder. In this paper, we propose a model with double encoders for 3D biomedical image segmentation. Our model is a U-shaped CNN augmented with an independent transformer encoder. We fuse the information from the convolutional encoder and the transformer, and pass it to the decoder to obtain the results. We evaluate our methods on three public datasets from three different challenges: BTCV, MoDA and Decathlon. Compared to the state-of-the-art models with and without transformers on each task, our proposed method obtains higher Dice scores across the board.