基于背景信息增强的宫颈异常细胞分类网络

Cervical Abnormal Cell Classification Network Based on Background Information Enhancement

摘要: 宫颈异常细胞的精准分类是宫颈癌早期辅助筛查的重要环节。针对传统方法在同一背景下异常细胞关联信息挖掘不足、上下文环境利用不充分以及图像尺度变化导致特征信息丢失等问题，本文提出一种基于背景信息增强的宫颈异常细胞分类网络。该方法引入图卷积增强同一背景下细胞之间的特征联系；设计背景特征感知机制，提升模型的上下文感知能力；引入多尺度双视野融合卷积，实现多尺度信息的有效整合。在多个数据集上的实验结果表明，该方法在分类准确率、召回率等指标上均优于多种主流方法，表现出良好的稳定性与鲁棒性，为宫颈癌早期自动化筛查提供了一种有效的技术方案。

Abstract: Accurate classification of cervical abnormal cells is a crucial step in the early auxiliary screening of cervical cancer. To address the limitations of traditional methods, including insufficient exploration of correlations among abnormal cells within the same background, inadequate utilization of contextual environmental information, and loss of feature information caused by image scale variations, this paper proposes a cervical abnormal cell classification network based on background information enhancement. The proposed method introduces a graph convolution mechanism to strengthen the feature relationships among cells within the same background. A background feature perception module is designed to improve the model’s ability to capture contextual information. In addition, a multi-scale dual-receptive-field fusion convolution module is employed to effectively integrate features at different scales. Experimental results on multiple datasets demonstrate that the proposed method outperforms several mainstream approaches in terms of classification accuracy and recall, showing good stability and robustness. The proposed approach provides an effective technical solution for automated early screening of cervical cancer.