DEW-Net:一个W形双编码器网络,用于病理性H&E图像分割的注意力融合机制
IEEE journal of biomedical and health informatics
|January 27, 2026
概括
一个新的DEW-Net模型在肺癌图像中准确地细分了编程的细胞死亡连接体1 (PD-L1) 表达. 这种方法通过有效地融合本地和全球特征来提高细分精度,优于现有的方法.
科学领域:
- 数字病理学数字病理学
- 医疗图像分析 医疗图像分析
- 计算瘤学是一种计算瘤学.
背景情况:
- 在肺状细胞癌 (LSCC) 中,精确细分编程细胞死亡配体1 (PD-L1) 表达对于治疗决策至关重要.
- 病理性H&E图像由于形态异质性和各种表达区域大小而存在挑战.
- 现有的混合CNN-变压器模型在有效的信息融合方面扎,限制了PD-L1细分精度.
研究的目的:
- 开发一种先进的深度学习模型,用于精确的像素级别对LSCC中的PD-L1表达的细分.
- 增强信息交互,减少在病理图像分析中的特征融合过程中的冗余数据.
- 提高PD-L1细分模型的准确性和概括能力.
主要方法:
- 提出了一个W形双编码器网络 (DEW-Net),并行集成一个CNN编码器和一个Swin变压器编码器.
- 引入了交叉注意力融合 (CAF) 模块,以改进语义特征融合和信息交互.
- 集成的通道注意 (CA) 和双边投票位置注意 (BPA) 模块来完善特征表示和减少噪音.
主要成果:
- 在四个数据集的PD-L1细分任务中,DEW-Net实现了卓越的性能.
- 该模型在PD-L1细分数据集上达到了79.93%的子相似系数 (DSC) 和71.27%的跨欧交点 (IoU).
- 在细分精度和概括方面,与最先进的 (SOTA) 方法相比显著改进.
结论:
- 拟议的DEW-Net有效地解决了LSCC病理图像中PD-L1细分的挑战.
- 新的注意力融合机制增强了特征交互,减少了冗余信息,从而提高了细分精度.
- 在精密瘤学和生物标志物分析中,DEW-Net具有很强的临床应用潜力.
相关概念视频
Nuclear Fusion
33.8K
The process of converting very light nuclei into heavier nuclei is also accompanied by the conversion of mass into large amounts of energy, a process called fusion. The principal source of energy in the sun is a net fusion reaction in which four hydrogen nuclei fuse and ultimately produce one helium nucleus and two positrons.
A helium nucleus has a mass that is 0.7% less than that of four hydrogen nuclei; this lost mass is converted into energy during the fusion. This reaction produces about...
A helium nucleus has a mass that is 0.7% less than that of four hydrogen nuclei; this lost mass is converted into energy during the fusion. This reaction produces about...
33.8K
Protein Networks
4.5K
An organism can have thousands of different proteins, and these proteins must cooperate to ensure the health of an organism. Proteins bind to other proteins and form complexes to carry out their functions. Many proteins interact with multiple other proteins creating a complex network of protein interactions.
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
These interactions can be represented through maps depicting protein-protein interaction networks, represented as nodes and edges. Nodes are circles that are representative of a protein,...
4.5K
Encoding
837
Information enters the brain through encoding, which is the input of information into the memory system. Once sensory information is received from the environment, the brain labels or codes it. The information is then organized with similar information and connected to existing concepts. Encoding occurs through automatic processing and effortful processing.
Automatic processing involves the encoding of details like time, space, frequency, and the meaning of words, usually done without conscious...
Automatic processing involves the encoding of details like time, space, frequency, and the meaning of words, usually done without conscious...
837
Molecular Shape and Polarity
75.4K
Dipole Moment of a Molecule
75.4K
Network Covalent Solids
16.1K
Network covalent solids contain a three-dimensional network of covalently bonded atoms as found in the crystal structures of nonmetals like diamond, graphite, silicon, and some covalent compounds, such as silicon dioxide (sand) and silicon carbide (carborundum, the abrasive on sandpaper). Many minerals have networks of covalent bonds.
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
To break or to melt a covalent network solid, covalent bonds must be broken. Because covalent bonds are relatively strong, covalent network solids are typically...
16.1K
VSEPR Theory and the Basic Shapes
84.4K
Overview of VSEPR Theory
84.4K


