论著摘要 |【AI-DR】使用深度学习的胸部计算机断层扫描对吸烟患者的疾病分期和预后(双语版)

2018-01-08 13:50:52 admin 0

Disease Staging and Prognosis in Smokers Using Deep Learning in Chest Computed Tomography.

发表日期: 2017.09.11   来源:Am J Respir Crit Care Med. 2017 Sep 11.

作者:

González G1, Ash SY2, Vegas Sanchez-Ferrero G3 , Onieva Onieva J4, Rahaghi FN5, Ross JC6, Díaz A7, San José Estépar R8, Washko GR9; COPDGene and ECLIPSE investigators.

作者介绍:

1. Sierra Research, Alicante, Spain ; german.gonzalez.serrano@gmail.com.

2. Brigham and Women\'s Hospital, 1861, Medicine, Boston, Massachusetts, United States ; syash@partners.org.

3. Brigham And Women's Hospital, Department of Radiology, Boston, Massachusetts, United States ; gvegas@bwh.harvard.edu.

4. Brigham and Women\'s Hospital, 1861, Laboratory of Mathematics in Imaging, Department of Radiology, Boston, Massachusetts, United States ; jonieva@bwh.harvard.edu.

5. Brigham and Women\'s Hospital, 1861, Medicine, Boston, Massachusetts, United States ; frahaghi@bwh.harvard.edu.

6. Brigham and Womens Hospital, Boston, Massachusetts, United States ; jross@bwh.harvard.edu.

7. Division of Pulmonary and Critical Care Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, United States ; ADIAZ6@PARTNERS.ORG.

8. Brigham and Women\'s Hospital, Radiology , Boston, Massachusetts, United States ; rsanjose@bwh.harvard.edu.

9. Brigham and Women's Hospital, Pulmonary and Critical Care Medicine, Boston, Massachusetts, United States ; gwashko@partners.org.

摘要

Abstact

基本原理

深度学习是一个强大的工具,可能允许改进的结果预测。

RATIONALE

Deep learning is a powerful tool that may allow for improved outcome prediction.

目的

为了确定深入学习,特别是卷积神经网络(CNN)分析,可以检测和分期慢性阻塞性肺疾病(COPD)并预测吸烟者的急性呼吸道疾病事件(ARD)和死亡率。

OBJECTIVES

To determine if deep learning, specifically convolutional neural network (CNN) analysis, could detect and stage chronic obstructive pulmonary disease (COPD) and predict acute respiratory disease events (ARD) and mortality in smokers.

方法

使用来自7,983名COPDGene参与者的CT扫描训练CNN,并使用1000名非重叠COPDGene参与者和1672名ECLIPSE参与者进行评估。Logistic回归(c统计和Hosmer-Lemeshow检验)用于评估COPD诊断和ARD预测。 Cox回归(c指数和Greenwood-Nam-D'Agnostino检验)用于评估死亡率。

METHODS

A CNN was trained using CT scans from 7,983 COPDGene participants and evaluated using 1000 non-overlapping COPDGene participants and 1,672 ECLIPSE participants. Logistic regression (c-statistic and the Hosmer-Lemeshow test) was used to assess COPD diagnosis and ARD prediction. Cox regression (c-index and the Greenwood-Nam-D'Agnostino test) was used to assess mortality.

测量和主要的实验结果

在COPDGene中,COPD检测的c统计量为0.856。COPDGene参与者的51.1%准确分期,74.95%在一个阶段。在ECLIPSE中,29.4%准确分期,74.6%在一个阶段。在COPDGene和ECLIPSE中,ARD事件的c统计分别为0.64和0.55,Hosmer-Lemeshow分别为0.502和0.380,表明没有校正差的证据。在COPDGene和ECLIPSE中,CNN预测死亡率具有公平的歧视(c指数分别为0.72和0.60),没有校正不足的证据(Greenwood-Nam-D'Agnostino p值分别为0.307和0.331)。

MEASUREMENTS AND MAIN RESULTS

In COPDGene, the c-statistic for the detection of COPD was 0.856. 51.1% of participants in COPDGene were accurately staged and 74.95% were within one stage. In ECLIPSE, 29.4% were accurately staged and 74.6% were within one stage. In COPDGene and ECLIPSE the c-statistics for ARD events were 0.64 and 0.55 respectively and the Hosmer-Lemeshow p=0.502 and 0.380 respectively, suggesting no evidence of poor calibration. In COPDGene and ECLIPSE, CNN predicted mortality with fair discrimination (c-indices 0.72 and 0.60 respectively), and without evidence of poor calibration (Greenwood-Nam-D'Agnostino p-values of 0.307 and 0.331 respectively).

结论

仅使用CT成像数据的深度学习方法可以识别那些具有COPD并且预测谁最有可能患有ARD事件的吸烟者和死亡率最高的吸烟者。在人口层面,CNN分析可能是风险评估的有力工具。

CONCLUSIONS

A deep learning approach that uses only CT imaging data can identify those smokers who have COPD and predict who are most likely to have ARD events and those with the highest mortality. At a population level CNN analysis may be a powerful tool for risk assessment.

关键词:

人工智能(计算机视觉系统);神经网络;慢性阻塞性肺疾病;X射线计算

Keywords:

Artificial Intelligence (Computer Vision Systems); Neural Networks; Pulmonary Disease, Chronic Obstructive; X-Ray Computed

阅读原文:PMID: 28892454  DOI: 10.1164/rccm.201705-0860OC



标签:   人工智能 AI 神经网络 CNN 慢性阻塞性肺疾病 卷积神经网络 cox回归 X射线计算

慧影医疗科技(北京)有限公司

地点:北京市海淀区中关村东升科技园B2-C103

电话:400-890-9020

邮箱:radcloud@huiyihuiying.com

关闭
图片
图片