使用转移学习与最先进的CNN架构进行疟疾寄生虫细胞分类
Azhar Ali Laghari1, Wazir Muhammad2, Mudasar Latif Memon3
1College of Resources and Environment, Shanxi Agricultural University, Taigu 030801, China.
Biology
|December 30, 2025
概括
深度学习模型,特别是ResNet-50和ResNet-101,在血液涂抹图像中自动检测疟疾寄生虫时显示出高准确度. 这种转移学习方法为疟疾诊断提供了比传统显微镜更快,更一致的替代方案.
科学领域:
- 医疗成像医学成像
- 计算生物学 计算生物学
- 寄生虫学的寄生虫学
背景情况:
- 疟疾诊断是关键的,但由于不可靠的显微镜方法和与其他发烧性疾病重叠的症状而具有挑战性.
- 不准确的疟疾诊断导致治疗延迟,增加严重并发症和死亡风险.
- 传统的显微镜诊断是劳动密集型的,需要专家技能,并遭受观察者之间的变化.
研究的目的:
- 研究深度学习的有效性,特别是预训练的卷积神经网络 (CNN) 模型,用于自动检测和分类疟疾寄生虫.
- 利用转移学习来克服诸如有限的标记数据等挑战,并加速疟疾检测模型的开发.
- 从显微镜图像进行疟疾诊断,比较各种最先进的CNN架构的性能.
主要方法:
- 使用了八个预训练的CNN模型 (VGG16,VGG19,Inception-v3,ResNet-18,ResNet-34,ResNet-50,ResNet-101,Xception) 来进行疟疾寄生虫的分类.
- 应用转移学习通过微调这些模型对显微镜血涂片图像的广泛标记数据集进行微调.
- 使用定量指标评估模型性能,包括精度,回忆,F1得分和准确性.
主要成果:
- ResNet-50和ResNet-101的准确率约为89%,而Xception的准确率约为88%.
- VGG-16 证明了精度回忆的权衡,实现了高精度,但对寄生细胞的回忆率较低,整体准确率约为 80%.
- ResNet-50,ResNet-101和Xception表现出强,平衡的性能,表明它们适合用于自动疟疾检测.
结论:
- 深度学习,特别是使用ResNet-50和ResNet-101等先进的CNN进行转移学习,为疟疾寄生虫检测提供了有效和准确的自动化解决方案.
- 拟议的深度学习方法比传统方法提供了显著的改进,承诺更一致和更有效的疟疾诊断.
- 这些发现支持人工智能工具的临床实用性,以提高疟疾诊断的准确性和及时性.
相关概念视频
Symbiosis
36.8K
Symbiotic relationships are long-term, close interactions between individuals of different species that affect the distribution and abundance of those species. When a relationship is beneficial to both species, this is called mutualism. When the relationship is beneficial to one species but neither beneficial nor harmful to the other species, this is called commensalism. When one organism is harmed to benefit another, the relationship is known as parasitism. These types of relationships often...
36.8K
Diversity of Protists II
745
Alveolates are a group of organisms recognized by the presence of alveoli, which are cytoplasmic sacs located beneath the cell membrane. While their function remains uncertain, alveoli may help regulate water balance by controlling how much water enters and leaves the cell. In dinoflagellates, these structures may serve as armor plates. There are three major types of alveolates: ciliates, which move using cilia; dinoflagellates, which use flagella for movement; and apicomplexans, which are...
745


