沈阳大学学报（自然科学版）2025，Vol.37Issue(2)：147-154,161,9.
融合傅里叶变换和可学习矩阵的肝脏肿瘤CT图像分割
Liver Tumor CT Image Segmentation Based on Integration of Fourier Transform and Learnable Matrices
邵虹 1张雨1
作者信息
- 1. 沈阳工业大学信息科学与工程学院,辽宁沈阳 110870
- 折叠
摘要
Abstract
To achieve precise automatic segmentation of liver tumors,a 3D deep U-shaped network incorporating residual attention was constructed to segment the liver,thereby reducing the impact of other organs in the background on tumor segmentation.Based on the liver segmentation,a gating mechanism that integrates Fourier transform and learnable matrices was proposed.This mechanism filters out irrelevant information while better handling edge regions,enhancing the model's ability to discern tumor edge information.The results demonstrate that the algorithm achieves excellent performance in liver tumor segmentation,with a dice similarity coefficient(DSC)of 78.99％,representing a 7.15％improvement over the 3D U-Net model.This enhances the segmentation accuracy of liver tumors and exhibits strong generalization capabilities.关键词
肝脏肿瘤分割/残差模块/注意力机制/傅里叶变换/可学习矩阵Key words
liver tumor segmentation/residual module/attention mechanism/Fourier transform/learnable matrix分类
计算机与自动化引用本文复制引用
邵虹,张雨..融合傅里叶变换和可学习矩阵的肝脏肿瘤CT图像分割[J].沈阳大学学报（自然科学版）,2025,37(2):147-154,161,9.
