半监督的多类肺炎分类使用CNN级联森林框架
P Muthukumaraswamy1, T Yuvaraj2, R Krishnamoorthy3
1Department of Biomedical Engineering, Kings Engineering College, Chennai, 602117, India. muthukumaraswamy@kingsedu.ac.in.
Scientific reports
|February 5, 2026
概括
这项研究引入了一种新型的半监督CNN增强级联森林 (CE-Cascade) 模型,用于从医学图像中准确的多类肺炎分类. 该CE-Cascade模型实现了高分类率,超过了现有的深度学习系统.
科学领域:
- 医疗成像中的人工智能
- 深度学习用于疾病诊断和诊断
- 计算病理学计算病理学
背景情况:
- 精确的医学成像诊断肺炎亚型至关重要,但受到有限的注释数据和成像变异性的挑战.
- 目前的深度学习方法通常是单变量和二进制分类,限制了对多类肺炎检测的临床适用性.
研究的目的:
- 提出一个半监督的CNN-增强布森林 (CE-布) 计划,以使用胸部X射线和CT图像进行强大的多类肺炎分类.
- 通过利用标记和未标记的数据来解决现有方法的局限性,以提高概括性.
主要方法:
- 一个卷积神经网络 (CNN) 从医疗图像中提取了深层特征.
- 一个级联森林改进了这些特征,以确定对肺炎检测具有重要意义的层次和多层次模式.
- 一个半监督的伪标记策略利用未标记的数据来增强模型的概括性.
主要成果:
- 评估了CE-Cascade框架的4578个胸部X射线和CT图像,包括细菌,病毒,真菌,一般性肺炎和正常类别.
- 该模型实现了98.86%的整体分类精度,超过了最先进的深度学习系统.
- 实验发现表明该模型在分类不同类型的肺炎方面的有效性.
结论:
- 将CNN和Cascade Forest与半监督学习相集成,为自动化多类肺炎分类提供了一个强大的方法.
- CE-Cascade方法提供了一种临床上有意义的解决方案,克服了稀缺注释和交叉成像变化的局限性.
- 这种自动化系统显示了改善医学成像诊断准确性的巨大潜力.
相关概念视频
01:24Intracellular Signaling Cascades
53.6K
Once a ligand binds to a receptor, the signal is transmitted through the membrane and into the cytoplasm. The continuation of a signal in this manner is called signal transduction. Signal transduction only occurs with cell-surface receptors, which cannot interact with most components of the cell, such as DNA. Only internal receptors can interact directly with DNA in the nucleus to initiate protein synthesis. When a ligand binds to its receptor, conformational changes occur that affect the...
53.6K
01:25Rab Cascades
3.6K
Rab GTPases act in a regulated cascade during membrane fusion, helping the lipid bilayers mix. The Rab family of proteins are active when bound to GTP, and inactive when bound to GDP. Hence, they act as guanine nucleotide-dependent molecular switches. Rab-GTP recognizes and binds to long or short-range tethering proteins to capture the target vesicle. These tethers coordinate with SNAREs on the vesicle and the target membrane to assemble the trans SNARE complex that locks the mixing bilayers.
3.6K
01:30Pneumonia I: Introduction
903
Pneumonia is an acute respiratory infection that targets the lungs, specifically the alveoli. These tiny air sacs, essential for oxygen exchange, become engorged with pus and fluid, severely hindering breathing, decreasing oxygen absorption, and causing significant pain and discomfort during respiration.
Risk Factors
Various factors influence the likelihood of developing pneumonia. Age plays a crucial role, with infants, children under two, and individuals over 65 at increased risk due to their...
Risk Factors
Various factors influence the likelihood of developing pneumonia. Age plays a crucial role, with infants, children under two, and individuals over 65 at increased risk due to their...
903
01:29Pneumonia II: Pathophysiology
3.0K
The pathophysiology of pneumonia involves the following steps:
3.0K
01:22Amplifying Signals via Enzymatic Cascade
18.6K
When a ligand binds to a cell-surface receptor, the receptor's intracellular domain changes shape, which may either activate its enzyme function or allow its binding to other molecules. The initial signal is amplified by most signal transduction pathways. This means that a single ligand molecule can activate multiple molecules of a downstream target. Proteins that relay a signal are most commonly phosphorylated at one or more sites, activating or inactivating the protein. Kinases catalyze...
18.6K
01:07MAPK Signaling Cascades
8.5K
Mitogen-activated protein kinase, or MAPK pathway, activates three sequential kinases to regulate cellular responses such as proliferation, differentiation, survival, and apoptosis. The canonical MAPK pathway starts with a mitogen or growth factor binding to an RTK. The activated RTKs stimulate Ras, which recruits Raf or MAP3 Kinase (MAPKKK), the first kinase of the MAPK signaling cascade. Raf further phosphorylates and activates MEK or MAP2 Kinases (MAPKK), which in turn phosphorylates MAP...
8.5K