论著摘要 |【PET】使用遗传算法和随机森林分类器在食管癌中预测结果的特征选择(双语版)

2017-08-22 14:19:52 admin 7

Feature selection for outcome prediction in oesophageal cancer using genetic algorithm and random forest classifier.

发表日期:2016.12.28    来源:Comput Med Imaging Graph.

作者:Paul D1Su R2Romain M3Sébastien V4Pierre V3Isabelle G3.

作者介绍

    1.LITIS - QUANTIF, University of Rouen, 22, boulevard Gambetta, 76000 Rouen, France; DOSISOFT, 45/47, avenue Carnot, 94230 Cachan, France.

    2.LITIS - QUANTIF, University of Rouen, 22, boulevard Gambetta, 76000 Rouen, France.

    3.LITIS - QUANTIF, University of Rouen, 22, boulevard Gambetta, 76000 Rouen, France; Henri Becquerel Centre, 1, rue d'Amiens, 76038 Rouen Cedex, France.

    4.DOSISOFT, 45/47, avenue Carnot, 94230 Cachan, France.

摘要

患者的预后预测可以大大有助于个性化癌症治疗。 大量的定量特征(临床检查,成像,...)可能有助于评估患者的预后。 挑战是选择最具预测性的一部分特征。 在本文中,我们提出了一种从正电子发射断层扫描(PET)图像和临床数据提取的新的特征选择策略,称为GARF(基于随机森林的遗传算法)。使用GARF选择了65位符合放化疗的局部晚期食管癌患者的可预测治疗反应或治疗结束3年后患者存活率的最相关特征。 通过9个特征的子集获得最相关的预测结果,导致随机森林错误分类率为18±4%,受试者工作特征(ROC)曲线下面积(AUC)为0.823±0.032。 8个特征得出最相关的预后结果,误差率为20±7%,AUC为0.750±0.108。 预测和预后结果都显示使用GARF比使用其他4种研究方法更好的表现。

The outcome prediction of patients can greatly help to personalize cancer treatment. A large amount of quantitative features (clinical exams, imaging, …) are potentially useful to assess the patient outcome. The challenge is to choose the most predictive subset of features. In this paper, we propose a new feature selection strategy called GARF (genetic algorithm based on random forest) extracted from positron emission tomography (PET) images and clinical data. The most relevant features, predictive of the therapeutic response or which are prognoses of the patient survival 3 years after the end of treatment, were selected using GARF on a cohort of 65 patients with a local advanced oesophageal cancer eligible for chemo-radiation therapy. The most relevant predictive results were obtained with a subset of 9 features leading to a random forest misclassification rate of 18±4% and an areas under the of receiver operating characteristic (ROC) curves (AUC) of 0.823±0.032. The most relevant prognostic results were obtained with 8 features leading to an error rate of 20±7% and an AUC of 0.750±0.108. Both predictive and prognostic results show better performances using GARF than using 4 other studied methods.

关键词

特征选择,遗传算法,食道癌,放射组学,随机森林

Feature selection; Genetic algorithm; Oesophageal cancer; Radiomics; Random forest


阅读原文:10.1016/j.compmedimag.2016.12.002


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
  • 人工智能诊断云平台