Multimodal Brain Tumor Classification Using Deep Learning and Robust Feature Selection: A Machine Learning Application for Radiologists.

Muhammad Attique Khan,Syed Ahmad Chan Bukhari,Majed Alhaisoni,Amjad Rehman,Imran Ashraf,Robertas Damaševičius,Rafal Scherer

doi:10.3390/diagnostics10080565

Abstract

Manual identification of brain tumors is an error-prone and tedious process for radiologists; therefore, it is crucial to adopt an automated system. The binary classification process, such as malignant or benign is relatively trivial; whereas, the multimodal brain tumors classification (T1, T2, T1CE, and Flair) is a challenging task for radiologists. Here, we present an automated multimodal classification method using deep learning for brain tumor type classification. The proposed method consists of five core steps. In the first step, the linear contrast stretching is employed using edge-based histogram equalization and discrete cosine transform (DCT). In the second step, deep learning feature extraction is performed. By utilizing transfer learning, two pre-trained convolutional neural network (CNN) models, namely VGG16 and VGG19, were used for feature extraction. In the third step, a correntropy-based joint learning approach was implemented along with the extreme learning machine (ELM) for the selection of best features. In the fourth step, the partial least square (PLS)-based robust covariant features were fused in one matrix. The combined matrix was fed to ELM for final classification. The proposed method was validated on the BraTS datasets and an accuracy of 97.8%, 96.9%, 92.5% for BraTs2015, BraTs2017, and BraTs2018, respectively, was achieved.

Highlights

A brain tumor is an abnormal growth of brain cells in an uncontrollable way [1,2]
The results are provided for two different pipeline procedures, namely; (i) feature extraction from fully connected (FC) layer seven and a performed feature selection approach that followed the feature fusion and classification and (ii) which followed the proposed architecture, as given in
It can be seen that, if the contrast enhancement step is not employed, the results show a decrease of almost 7% of accuracy for all BraTS datasets

Summary

Introduction

A brain tumor is an abnormal growth of brain cells in an uncontrollable way [1,2]. Brain tumors can be cancerous or noncancerous. The gravity inside the skull can accelerate the growth of a brain tumor. Various manifestations and classes of brain tumors have different appearances on magnetic resonance imaging (MRI) data [4,5]. MRI scans are typically used to detect and classify brain tumors. MRI assists doctors in evaluating tumors in order to plan for further treatment. This treatment depends on various factors like shape, size, type, grade, and location of cancer. Accurate recognition and classification of brain tumors are critical for proper treatment [6]

Methods

Results

Discussion

Conclusion