Machine Learning and Deep Learning Techniques to Predict Overall Survival of Brain Tumor Patients Using MRI Images

Document Type

Article

Publication Date

1-1-2018

Publication Title

2017 IEEE 17th International Conference on Bioinformatics and Bioengineering (BIBE)

Volume

2017

First page number:

9

Last page number:

14

Abstract

This paper presents a method to automatically predict the survival rate of patients with a glioma brain tumor by classifying the patients MRI image using machine learning (ML) methods. The dataset used in this study is BraTS 2017, which provides 163 samples; each sample has four sequences of MRI brain images, the overall survival time in days, and the patients age. The dataset is labeled into three classes of survivors: short-term, mid-term, and long-term. To improve the prediction results, various types of features were extracted and trained by various ML methods. Features considered included volumetric, statistical and intensity texture, histograms and deep features; ML techniques employed included support vector machine (SVM), k-nearest neighbors (KNN), linear discriminant, tree, ensemble and logistic regression. The best prediction accuracy based on classification is achieved by using deep learning features extracted by a pre-trained convolutional neural network (CNN) and was trained by a linear discriminant.

Keywords

Convolutional neural network; Deep learning features; Discrete wavelet transform; Gliomagray level co-occurrence matrix; Intensity texture features; K-nearest neighbors; Logistic regression; Pre-trained CNN; Statistical texture features; Support vector machine; Volumetric features

Disciplines

Biomedical Engineering and Bioengineering

Language

English

Find It

UNLV article access

Search your library

Share

COinS