CoT-UNet++: A medical image segmentation method based on contextual transformer and dense connection

Abstract

<abstract> <p>Accurate depiction of individual teeth from CBCT images is a critical step in the diagnosis of oral diseases, and the traditional methods are very tedious and laborious, so automatic segmentation of individual teeth in CBCT images is important to assist physicians in diagnosis and treatment. TransUNet has achieved success in medical image segmentation tasks, which combines the advantages of Transformer and CNN. However, the skip connection taken by TransUNet leads to unnecessary restrictive fusion and also ignores the rich context between adjacent keys. To solve these problems, this paper proposes a context-transformed TransUNet++ (CoT-UNet++) architecture, which consists of a hybrid encoder, a dense connection, and a decoder. To be specific, a hybrid encoder is first used to obtain the contextual information between adjacent keys by CoTNet and the global context encoded by Transformer. Then the decoder upsamples the encoded features by cascading upsamplers to recover the original resolution. Finally, the multi-scale fusion between the encoded and decoded features at different levels is performed by dense concatenation to obtain more accurate location information. In addition, we employ a weighted loss function consisting of focal, dice, and cross-entropy to reduce the training error and achieve pixel-level optimization. Experimental results demonstrate that the proposed CoT-UNet++ method outperforms the baseline models and can obtain better performance in tooth segmentation.</p> </abstract>

Deep learning-based tooth segmentation methods in medical imaging: A review

“…Yin et al 23 proposed CoT-UNet++, an improved network based on the TransUNet 76 network architecture, which consists of a hybrid encoder, dense connections, and a decoder. CoT-UNet++ uses a hybrid encoder to obtain neighboring context information for CoTNet 77 coding and global context for Transformer coding.…”

“…99 Transformer was initially the top network model in natural language processing (NLP), and has also shown excellent performance when it applied to image tasks. Several works 23,24 have used the Transformer attention mechanism to achieve better tooth segmentation.…”

“…Transformer takes full advantage of the multi-head attention mechanism, which achieves a performance improvement over many previous deep learning models based on CNNs. Some works [22][23][24] realized tooth segmentation with transformer as the backbone module. Xiong et al 24 achieved a 96% dice score on a private dataset containing 16,000 3D dental models.…”

Improved UNet with Attention for Medical Image Segmentation

“…Wang et al [8] argued that TransUNet, similar to the UNet architectures, has skip connections at the same level, which constrains feature fusion. Instead, new model architectures such as MS-TransUNet++ [8] and CoT-UNet++ have been proposed [47]. These models use dense skip connections between the encoder and decoder at different levels to improve feature fusion, similar to UNet++, in addition to a hybrid encoder that includes a Transformer.…”