通过深度学习进行染色体分类,并将其应用于染色体结构异常的患者
Chuan Yang1, Tingting Li2, Qiulei Dong3
1Department of Clinical Genetics, Shengjing Hospital of China Medical University, Shenyang 110004, China; Department of Cardiology, Shengjing Hospital of China Medical University, Shenyang 110004, China.
Medical engineering & physics
|November 20, 2023
概括
一个新的深度卷积神经网络 (DCNN) 模型自动化了用于遗传疾病诊断的染色体分类. 这种人工智能工具提高了 karyotyping 的准确性和效率,有助于临床遗传选.
科学领域:
- 遗传学和基因组学 遗传学和基因组学
- 计算生物学 计算生物学
- 医学诊断 医学诊断 医学诊断
背景情况:
- 型鉴定对于诊断遗传疾病至关重要,但通常是手动的,缓慢的,容易出现错误.
- 目前的临床型化方法在时间和准确性方面存在重大挑战.
研究的目的:
- 开发一种用于染色体分类的自动化,单阶段深卷积神经网络 (DCNN) 模型.
- 为了实现正常和异常染色体的端到端分类,提高诊断效率.
主要方法:
- 分析了2,424个正常和544个异常染色体的数据集.
- 一个初步的支持矢量机 (SVM) 模型被用于基线性能评估.
- 开发了一个深度卷积神经网络 (DCNN) 模型,并应用于相同的数据集.
主要成果:
- DCNN模型在分类24个正常染色体方面实现了91.75%的准确性,超过了SVM的86.01%的性能.
- 对于32种正常和异常染色体,DCNN模型的准确率达到87.76%,而SVM的准确率为85.37%.
- 区分八种常见结构异常的准确度在90.84%至100%之间,AUC值在91.81%至100%之间.
结论:
- 拟议的DCNN模型为型分类提供了一个有效的端到端解决方案.
- 这种人工智能工具可以作为一种有价值的预测工具,用于检测异常的型,而无需手动的特征提取.
- 该DCNN模型有潜力降低成本,提高临床遗传选和诊断的效率.
相关概念视频
Karyotyping
61.3K
Overview
61.3K
Lampbrush Chromosomes
7.9K
In 1882, Flemming observed lampbrush chromosomes (LBC) in salamander eggs. Later in 1892, Rückert observed LBCs in shark egg cells and coined the term "lampbrush chromosomes" because they looked like brushes used to clean kerosene lamps.
LBCs are made up of two pairs of conjugating homologous chromatids. Each chromatid consists of alternatively positioned regions of condensed-inactive chromatin and loosely placed-active side loops, which can be contracted and extended. The loops...
LBCs are made up of two pairs of conjugating homologous chromatids. Each chromatid consists of alternatively positioned regions of condensed-inactive chromatin and loosely placed-active side loops, which can be contracted and extended. The loops...
7.9K
Polytene Chromosomes
3.0K
3.0K


