Xinyi Li1, Fei Jiang2, Yi Guo3, Zhendong Jin2, Yuanyuan Wang4. 1. Department of Electronic Engineering, Fudan University, Shanghai, China. 2. Department of Gastroenterology, Changhai Hospital, Shanghai, China. 3. Department of Electronic Engineering, Fudan University, Shanghai, China. guoyi@fudan.edu.cn. 4. Department of Electronic Engineering, Fudan University, Shanghai, China. yywang@fudan.edu.cn.
Abstract
PURPOSE: The purpose of our study is to propose a preoperative computer-aided diagnosis system based on a radiomics method to differentiate gastrointestinal stromal tumors (GISTs) of the higher-risk group (HRG) from those of the lower-risk group (LRG) on endoscopic ultrasound (EUS) images. MATERIALS AND METHOD: Gastro-EUS (G-EUS) images of four different risk level GISTs were collected from 19 hospitals. The datasheet included 168 case HRG GISTs and 747 case LRG GISTs. A radiomics method with image segmentation, feature extraction, feature selection and classification was developed. Here 439 radiomics features were firstly extracted, and then, the least absolute shrinkage selection operator (lasso) model with a tenfold cross-validation and 31 bootstraps was used to reduce the dimension of feature sets. Finally, random forest was applied to establish the classification model. RESULTS: The proposed model differentiated 32 case HRG GISTs from 149 case LRG GISTs. Result for the testing set achieved the area under the receiver operating characteristic curve of 0.839, the accuracy of 0.823, the sensitivity of 0.813 and the specificity of 0.826. CONCLUSION: The model could increase preoperative diagnostic accuracy and provide a valuable reference for the doctors.
PURPOSE: The purpose of our study is to propose a preoperative computer-aided diagnosis system based on a radiomics method to differentiate gastrointestinal stromal tumors (GISTs) of the higher-risk group (HRG) from those of the lower-risk group (LRG) on endoscopic ultrasound (EUS) images. MATERIALS AND METHOD: Gastro-EUS (G-EUS) images of four different risk level GISTs were collected from 19 hospitals. The datasheet included 168 case HRG GISTs and 747 case LRG GISTs. A radiomics method with image segmentation, feature extraction, feature selection and classification was developed. Here 439 radiomics features were firstly extracted, and then, the least absolute shrinkage selection operator (lasso) model with a tenfold cross-validation and 31 bootstraps was used to reduce the dimension of feature sets. Finally, random forest was applied to establish the classification model. RESULTS: The proposed model differentiated 32 case HRG GISTs from 149 case LRG GISTs. Result for the testing set achieved the area under the receiver operating characteristic curve of 0.839, the accuracy of 0.823, the sensitivity of 0.813 and the specificity of 0.826. CONCLUSION: The model could increase preoperative diagnostic accuracy and provide a valuable reference for the doctors.
Authors: Christopher D M Fletcher; Jules J Berman; Christopher Corless; Fred Gorstein; Jerzy Lasota; B Jack Longley; Markku Miettinen; Timothy J O'Leary; Helen Remotti; Brian P Rubin; Barry Shmookler; Leslie H Sobin; Sharon W Weiss Journal: Hum Pathol Date: 2002-05 Impact factor: 3.466
Authors: Martin E Blackstein; Jean-Yves Blay; Christopher Corless; David K Driman; Robert Riddell; Denis Soulières; Carol J Swallow; Shailendra Verma Journal: Can J Gastroenterol Date: 2006-03 Impact factor: 3.522
Authors: Hugo J W L Aerts; Emmanuel Rios Velazquez; Ralph T H Leijenaar; Chintan Parmar; Patrick Grossmann; Sara Carvalho; Sara Cavalho; Johan Bussink; René Monshouwer; Benjamin Haibe-Kains; Derek Rietveld; Frank Hoebers; Michelle M Rietbergen; C René Leemans; Andre Dekker; John Quackenbush; Robert J Gillies; Philippe Lambin Journal: Nat Commun Date: 2014-06-03 Impact factor: 14.919