Abstract: Chromosome karyotyping is of great significance in prenatal screening and diagnosis of genetic diseases. Chromosome classification models were generally limited by weak feature extraction, high sensitivity to image quality, and inadequate attention to local details. As a result, overall accuracy remained low, and short chromosomes were frequently misidentified. Therefore, a cascade classification framework with coarse-to-fine granularity, integrating image texture enhancement and super-resolution technology was proposed. Firstly, chromosomes were roughly classified based on the International System for Human Cytogenetic Nomenclature (ISCN) and divided into long chromosomes and short chromosomes to mitigate class imbalance and feature ambiguity. Secondly, for the long chromosome classification task, a feature enhancement module was added to optimize the classification model's ability to perceive the details of long chromosomes. Then, according to the characteristics of short chromosome images, super-resolution techniques were applied to improve image quality and model perception ability. The framework was evaluated on a private dataset, achieving an overall classification accuracy of 98.91% and an F1 score of 98.77%, with 99.01% for long chromosomes and 98.31% for short chromosomes. These results demonstrate that the cascade framework, through differentiated strategies and task-specific models, significantly enhances classification accuracy and model robustness.

Key words: chromosome cascade classification, feature enhancement, texture enhancement module, image super-resolution, channel attention mechanism

摘要： 摘 要: 染色体核型化分析在产前筛查和遗传疾病诊断中具有重要意义。现有染色体分类模型普遍存在特征提取能力不足、图像质量敏感性强以及对局部细节感知不充分等问题，导致染色体整体分类准确率不高，尤其是短小染色体常常被错误识别因此，提出了一种融合图像纹理增强与超分辨率技术的染色体粗粒度-细粒度级联分类框架。首先，基于国际人类细胞遗传命名系统（ISCN）对染色体进行粗分类，将其划分为长染色体与短小染色体组，减少类别不均衡与特征混淆问题。其次，针对长染色体分类任务，通过增加特征增强模块，以优化分类模型对长染色体的细节感知能力。然后，针对短小染色体图像特点，框架引入图像超分辨率技术以提升图像质量与模型感知能力。最后，在私有数据集上进行实验验证，染色体整体分类准确率达到98.91%，F1分数达到98.77%，其中长染色体的F1分数为99.01%，短小染色体的F1分数为98.31%。实验结果表明，染色体级联分类框架针对不同分类任务采取差异化分类策略与分类模型，显著提升了染色体分类精度和模型鲁棒性。

关键词: 关键词: 染色体级联分类, 特征增强, 纹理增强模块, 图像超分辨率, 通道注意力机制