Comparison of Feature Selection Methods and Machine Learning Classifiers for Radiomics Analysis in Glioma Grading

Pan Sun,Lin Shi,Defeng Wang,Vincent Ct Mok

doi:10.1109/access.2019.2928975

Pan Sun, Lin Shi + Show 2 more

Open Access

https://doi.org/10.1109/access.2019.2928975

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2019
Citations: 75	License type: CC BY 4.0

Affiliation: Chinese University of Hong Kong, Beihang University

Abstract

Radiomics-based researches have shown predictive abilities with machine-learning approaches. However, it is still unknown whether different radiomics strategies affect the prediction performance. The aim of this study was to compare the prediction performance of frequently utilized radiomics feature selection and classification methods in glioma grading. Quantitative radiomics features were extracted from tumor regions in 210 Glioblastoma (GBM) and 75 low-grade glioma (LGG) MRI subjects. Then, the diagnostic performance of sixteen feature selection and fifteen classification methods were evaluated by using two different test modes: ten-fold cross-validation and percentage split. Balanced accuracy and area under the curve (AUC) of the receiver operating characteristic were used to evaluate prediction performance. In addition, the roles of the number of selected features, feature type, MRI modality, and tumor sub-region were compared to optimize the radiomics-based prediction. The results indicated that the combination of feature selection method L 1 -based linear support vector machine (L 1 -SVM) and classifier multi-layer perceptron (MLPC) achieved the best performance in the differentiation of GBM and LGG in both ten-fold cross validation (balanced accuracy:0.944, AUC:0.986) and percentage split (balanced accuracy:0.953, AUC:0.981). For radiomics feature extraction, the enhancing tumor region (ET) combined with necrotic and non-enhancing tumor (NCR/NET) regions in T1 post-contrast (T1-Gd) modality provided more considerable tumor-related phenotypes than other combinations of tumor region and MRI modality. Our comparative investigation indicated that both feature selection methods and machine learning classifiers affected the predictive performance in glioma grading. Also, the cross-combination strategy for comparison of radiomics feature selection and classification methods provided a way of searching optimal machine learning model for future radiomics-based prediction.

Highlights

Glioma is the most common primary intracranial tumor in adults [1]
CLASSIFICATION METHODS We investigated fifteen machine-learning classifiers: gaussion naïve bayes (GNB), multinomial naï ve bayes (MNB), bernoulli naïve bayes (BNB), k-nearest neighborhood (KNN), random forest (RF), bagging (BAG), decision tree (DT), gradient boosting decision tree (GBDT), adaptive boosting (Adaboost), xgboost (XGB), linear discriminant analysis (LDA), logistic regression (LGR), linear support vector machine (Linear-SVM), radial basis function support vector machine (RBF-SVM) and multi-layer perceptron (MLPC)
The diagnostic performance was quantified by balanced accuracy and area under the curve (AUC)

Summary

Introduction

Glioma is the most common primary intracranial tumor in adults [1]. It might occur anywhere in brain and appear highly spatial-temporal heterogeneity. According to WHO, glioma can be classified into grades I-IV based on histologically malignant behavior [2]. Low-grade glioma (LGG, grades I and II) patients typically have more than five years. Survival whereas only 3-5% of glioblastoma (GBM, grade IV) patients survive more than five years, with median survival about 12 months [3]. GBM is the most common glioma histology type, accounting for 70% of primary brain tumors. Preoperative glioma grading, the differentiation between GBM and LGG, is of great importance for making diagnostic decisions in clinical [4]

Objectives

Methods

Results

Discussion

Conclusion