人工智能培训

搜索

人工智能论文:走向无监督的癌症分型:使用组织学视觉词典预测预后(Towards Unsupervised Cancer Subtyping: Predictin

[复制链接]
luolei 发表于 2019-3-15 13:24:40 | 显示全部楼层 |阅读模式
luolei 2019-3-15 13:24:40 161 0 显示全部楼层
人工智能论文:走向无监督的癌症分型:使用组织学视觉词典预测预后(Towards Unsupervised Cancer Subtyping: Predicting Prognosis Using A  Histologic Visual Dictionary)与常见的癌症不同,例如前列腺癌和乳腺癌,罕见癌症的肿瘤分级很难并且在很大程度上是不确定的,因为样本量小,完成这项任务所需的时间非常庞大,以及提取人类观察模式的固有困难。其中一个最具挑战性的例子是肝内胆管细胞癌(ICC),这是一种由胆道系统引起的原发性肝癌,其中存在公认的肿瘤异质性,没有分级范例或预后生物标志物。在本文中,我们提出了一种新的无监督深度卷积自动编码器的聚类模型,该模型基于视觉相似性将246个ICC数字化整个幻灯片中的细胞和结构形态组合在一起。从这个组织学模式的可视字典中,我们使用聚类作为协变量来构建Cox比例危险生存模型。在单变量分析中,三个群体与无复发生存率显着相关。这些群集的组合在多变量分析中是显着的。在所有聚类的多变量分析中,五个显示无复发生存的显着性,但是整体模型未被测量为显着的。最后,病理学家将临床术语分配到视觉词典中的重要聚类,并找到支持这一假设的证据,即富含胶原蛋白的纤维化在疾病严重程度中发挥作用。这些结果提供了对癌症亚型分类未来的见解,并表明计算机病理学可以促进疾病预测。 ,尤其是罕见的癌症。
Unlike common cancers, such as those of the prostate and breast, tumorgrading in rare cancers is difficult and largely undefined because of smallsample sizes, the sheer volume of time needed to undertake on such a task, andthe inherent difficulty of extracting human-observed patterns.One of the mostchallenging examples is intrahepatic cholangiocarcinoma (ICC), a primary livercancer arising from the biliary system, for which there is well-recognizedtumor heterogeneity and no grading paradigm or prognostic biomarkers.In thispaper, we propose a new unsupervised deep convolutional autoencoder-basedclustering model that groups together cellular and structural morphologies oftumor in 246 ICC digitized whole slides, based on visual similarity.From thisvisual dictionary of histologic patterns, we use the clusters as covariates totrain Cox-proportional hazard survival models.In univariate analysis, threeclusters were significantly associated with recurrence-free survival.Combinations of these clusters were significant in multivariate analysis.In amultivariate analysis of all clusters, five showed significance torecurrence-free survival, however the overall model was not measured to besignificant.Finally, a pathologist assigned clinical terminology to thesignificant clusters in the visual dictionary and found evidence supporting thehypothesis that collagen-enriched fibrosis plays a role in disease severity.These results offer insight into the future of cancer subtyping and show thatcomputational pathology can contribute to disease prognostication, especiallyin rare cancers.人工智能论文:走向无监督的癌症分型:使用组织学视觉词典预测预后(Towards Unsupervised Cancer Subtyping: Predicting Prognosis Using A  Histologic Visual Dictionary) NEzCM0kjK0FEs3e3.jpg
URL地址:https://arxiv.org/abs/1903.05257     ----pdf下载地址:https://arxiv.org/pdf/1903.05257    ----人工智能论文:走向无监督的癌症分型:使用组织学视觉词典预测预后(Towards Unsupervised Cancer Subtyping: Predicting Prognosis Using A  Histologic Visual Dictionary)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

luolei当前离线
新手上路

查看:161 | 回复:0

快速回复 返回顶部 返回列表