| 吴源文,柳雪飞,蒋越,张宇涵.基于KL粗糙散度的医学图像多阈值分割方法[J].井冈山大学自然版,2025,46(6):71-78 |
| 基于KL粗糙散度的医学图像多阈值分割方法 |
| Multi-threshold segmentation method of medical image based on KL rough divergence |
| 投稿时间：2025-04-16 修订日期：2025-05-21 |
| DOI：10.3969/j.issn.1674-8085.2025.06.008 |
| 中文关键词: KL散度 粗糙集理论 KL粗糙散度 医学图像多阈值分割 |
| 英文关键词: KL divergence rough set theory KL rough divergence multi-threshold medical image segmentation |
| 基金项目:国家自然科学基金项目(62461004)；广西壮族自治区级大学生创新创业项目(S202410604066) |
|
| 摘要点击次数: 65 |
| 全文下载次数: 39 |
| 中文摘要: |
| 现有的多阈值优化方法普遍存在计算复杂度与分割精度的矛盾关系，其症结在于传统散度度量难以有效描述医学图像固有的模糊性和区域异质性。本研究创新性地构建了基于 KL 粗糙散度的多阈值分割模型，通过融合粗糙集近似空间理论，建立不确定性量化机制，借助粗糙集上下近似运算解析区域边界的模糊性特征。在优化策略上，引入改进型灰狼优化算法，通过设计动态收敛因子和贡献度占比策略，实现阈值空间的全局搜索与局部寻优的平衡。实验结果显示，该算法在 Dice 系数、Jaccard 数、敏感性和特异性等评估指标上均优于其他对比算法，表明该模型在分割精度和准确性方面具有显著优势。 |
| 英文摘要: |
| Existing multi-threshold optimization approaches generally exhibit the contradictory relationship between computational complexity and segmentation accuracy, its crux of the problem lies in that traditional divergence measures are difficult to effectively describe the inherent fuzziness and regional heterogeneity of medical images. The research innovatively constructs a multi-threshold segmentation model based on kullback-leibler rough divergence (KL rough divergence). By integrating rough set approximation space theory, the model establishes an uncertainty quantification mechanism: utilizing upper and lower approximation operations of rough sets to analyze fuzzy characteristics at regional boundaries. Regarding optimization strategies, an improved grey wolf optimizer (GWO) is introduced, achieving a balance between global search and local optimization in threshold space through designed dynamic convergence factors and contribution ratio strategies. The experimental results demonstrate that the proposed algorithm outperforms other comparative algorithms across evaluation metrics including coefficient, sensitivity, and specificity, indicating its significant advantages in segmentation precision and accuracy. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
