Improved UNet with Attention for Medical Image Segmentation

Abstract

Medical image segmentation is crucial for medical image processing and the development of computer-aided diagnostics. In recent years, deep Convolutional Neural Networks (CNNs) have been widely adopted for medical image segmentation and have achieved significant success. UNet, which is based on CNNs, is the mainstream method used for medical image segmentation. However, its performance suffers owing to its inability to capture long-range dependencies. Transformers were initially designed for Natural Language Processing (NLP), and sequence-to-sequence applications have demonstrated the ability to capture long-range dependencies. However, their abilities to acquire local information are limited. Hybrid architectures of CNNs and Transformer, such as TransUNet, have been proposed to benefit from Transformer’s long-range dependencies and CNNs’ low-level details. Nevertheless, automatic medical image segmentation remains a challenging task due to factors such as blurred boundaries, the low-contrast tissue environment, and in the context of ultrasound, issues like speckle noise and attenuation. In this paper, we propose a new model that combines the strengths of both CNNs and Transformer, with network architectural improvements designed to enrich the feature representation captured by the skip connections and the decoder. To this end, we devised a new attention module called Three-Level Attention (TLA). This module is composed of an Attention Gate (AG), channel attention, and spatial normalization mechanism. The AG preserves structural information, whereas channel attention helps to model the interdependencies between channels. Spatial normalization employs the spatial coefficient of the Transformer to improve spatial attention akin to TransNorm. To further improve the skip connection and reduce the semantic gap, skip connections between the encoder and decoder were redesigned in a manner similar to that of the UNet++ dense connection. Moreover, deep supervision using a side-output channel was introduced, analogous to BASNet, which was originally used for saliency predictions. Two datasets from different modalities, a CT scan dataset and an ultrasound dataset, were used to evaluate the proposed UNet architecture. The experimental results showed that our model consistently improved the prediction performance of the UNet across different datasets.

PIDNET: Polar Transformation Based Implicit Disentanglement Network for Truncation Artifacts

“…First, there is room for improvement in restoring edge details. In future studies, we can enhance the performance of the model by integrating novel techniques, such as attention mechanisms [39,40], into our approach. Secondly, although our model is unsupervised, it still requires the support of non-truncated data with similar features.…”

Enhancing bladder cancer diagnosis through transitional cell carcinoma polyp detection and segmentation: an artificial intelligence powered deep learning solution

“…Proposed by Xu et al (2021) for classifying COVID-19 positive cases from chest X-ray images, the MANet architecture incorporates attention mechanisms to enhance network capabilities and performance. The attention block directs the model’s focus towards critical areas of the input image, ensuring that irrelevant features do not interfere with the training process—an essential capability for medical image analysis ( Al Qurri and Almekkawy, 2023 ).…”

RCEAU-Net: Cascade Multi-Scale Convolution and Attention-Mechanism-Based Network for Laser Beam Target Image Segmentation with Complex Background in Coal Mine

“…Constrained by the traditional algorithm design difficulties, whereby it is difficult to ensure real-time target segmentation in the complex background of poor results and other issues, researchers have begun to choose to use deep learning-based image semantic segmentation methods to build segmentation models to complete the target segmentation task. Currently, the design of segmentation models based on the encoder–decoder structure of a full convolutional neural network FCN [ 9 ] is quite extensive, among which, due to the relatively simple structure of the U-Net [ 10 ] model and its outstanding segmentation performance, it and its variants have now achieved remarkable results in the semantic segmentation tasks of images such as medicine [ 11 ], traffic [ 12 ], agriculture [ 13 ], aerial photography [ 14 ], remote sensing [ 15 ], and so on. O. Oktay et al [ 16 ] proposed a novel Attention Gate (AG) model for the medical image domain, which can automatically learn to focus on target structures of different shapes and sizes, and integrated it into the U-Net network architecture to build the Attention U-Net network, which reduces the computational overheads of the original U-Net model, and improves the model’s sensitivity and computational accuracy.…”