Article Search
닫기

Original Article

Split Viewer

International Journal of Fuzzy Logic and Intelligent Systems 2021; 21(4): 349-357

Published online December 25, 2021

https://doi.org/10.5391/IJFIS.2021.21.4.349

© The Korean Institute of Intelligent Systems

DNN-Based Brain MRI Classification Using Fuzzy Clustering and Autoencoder Features

Nishant Chauhan* and Byung-Jae Choi*

Department of Electronic Engineering, Daegu University, Gyeongsan, Korea

Correspondence to :
Byung-Jae Choi (bjchoi@daegu.ac.kr)
*These authors contributed equally to this work.

Received: June 16, 2021; Revised: October 6, 2021; Accepted: December 22, 2021

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Worldwide interest has been noted in medical image analysis and classification using machine learning techniques. Magnetic resonance imaging (MRI) is one of the safe and painless procedures for human brain scanning. During the MRI procedure, magnetic fields and radio waves are used to scan and map the extended view of brain tissues for further pathological processes and analysis. For a qualitative and quantitative MRI analysis, the manual capability of radiologists and/or doctors is limited and time-consuming in complex and group-level diagnoses. Hence, the development of an intelligent, robust, and reliable support system for the diagnosis of brain-related diseases is a top priority. In this paper, a new deep neural network based MRI image classification approach is proposed that uses fuzzy c-mean (FCM) and an autoencoder to classify brain MRI as normal or abnormal, diminishing human error during the diagnosis of diseases in MRI scans. Here, FCM is utilized for abnormal tissue segmentation from brain MRI images, followed by an autoencoder, for extraction and dimensionality reduction of features. Finally, a deep neural network was used for the classification of brain MRI images that were trained using FCM-extracted features and sample data. Considering the availability of raw MRI data, data augmentation techniques have also been used to increase the number of data required to train a deep neural network. The experiment results achieved 96% accuracy and a 95% sensitivity rate for classification. The results demonstrate that the proposed well-trained deep learning technology has the potential to make solid predictions regarding brain abnormalities; therefore, it can be used as a prominent tool in clinical practice.

Keywords: Magnetic resonance imaging (MRI), Fuzzy c-mean (FCM) clustering, Deep neural network, Autoencoder, Classification

This research was partly supported by a Daegu University and Korea Institute of Advancement of Technology (KIAT) grant funded by the Korean government (MOTIE) (P0012724) and the Competency Development Program for Industry Specialists.

No potential conflict of interest relevant to this article was reported.

Nishant Chauhan received his B.S. degree in computer science from AKTU, India in 2012, and an M.S. degree in control and instrumentation from Daegu University, Korea in 2020. He is currently a full-time Ph.D. student in control and instrumentation at Daegu University, Department of Electronic Engineering (Graduate School). His research interests include intelligent control, fuzzy logic, image processing, machine learning, and deep learning for medical images, brain MRI, and fMRI analysis.

E-mail: nishantsep1090@daegu.ac.kr


Byung-Jae Choi received his B.S. degree in electronic engineering in 1987 from Kyungpook National University, Daegu. He received his M.S. and a Ph.D. degrees in electrical and electronic engineering, in 1989 and 1998, respectively, at KAIST, Daejeon, Korea. He has been a professor at the School of Electronic and Electrical Engineering, Daegu University, Daegu, Korea, since 1999. His current research interests include intelligent control and applications.

E-mail: bjchoi@daegu.ac.kr


Article

Original Article

International Journal of Fuzzy Logic and Intelligent Systems 2021; 21(4): 349-357

Published online December 25, 2021 https://doi.org/10.5391/IJFIS.2021.21.4.349

Copyright © The Korean Institute of Intelligent Systems.

DNN-Based Brain MRI Classification Using Fuzzy Clustering and Autoencoder Features

Nishant Chauhan* and Byung-Jae Choi*

Department of Electronic Engineering, Daegu University, Gyeongsan, Korea

Correspondence to:Byung-Jae Choi (bjchoi@daegu.ac.kr)
*These authors contributed equally to this work.

Received: June 16, 2021; Revised: October 6, 2021; Accepted: December 22, 2021

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Worldwide interest has been noted in medical image analysis and classification using machine learning techniques. Magnetic resonance imaging (MRI) is one of the safe and painless procedures for human brain scanning. During the MRI procedure, magnetic fields and radio waves are used to scan and map the extended view of brain tissues for further pathological processes and analysis. For a qualitative and quantitative MRI analysis, the manual capability of radiologists and/or doctors is limited and time-consuming in complex and group-level diagnoses. Hence, the development of an intelligent, robust, and reliable support system for the diagnosis of brain-related diseases is a top priority. In this paper, a new deep neural network based MRI image classification approach is proposed that uses fuzzy c-mean (FCM) and an autoencoder to classify brain MRI as normal or abnormal, diminishing human error during the diagnosis of diseases in MRI scans. Here, FCM is utilized for abnormal tissue segmentation from brain MRI images, followed by an autoencoder, for extraction and dimensionality reduction of features. Finally, a deep neural network was used for the classification of brain MRI images that were trained using FCM-extracted features and sample data. Considering the availability of raw MRI data, data augmentation techniques have also been used to increase the number of data required to train a deep neural network. The experiment results achieved 96% accuracy and a 95% sensitivity rate for classification. The results demonstrate that the proposed well-trained deep learning technology has the potential to make solid predictions regarding brain abnormalities; therefore, it can be used as a prominent tool in clinical practice.

Keywords: Magnetic resonance imaging (MRI), Fuzzy c-mean (FCM) clustering, Deep neural network, Autoencoder, Classification

Fig 1.

Figure 1.

Brain MRI scans: (a) normal brain, (b) brain tumor, and (c) brain damaged by Alzheimer’s disease.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 2.

Figure 2.

Types of brain MRI images: (a) FLAIR, (b) T1, (c) T2, (d) DWI, and (e) fMRI.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 3.

Figure 3.

Proposed model architecture based on deep learning approach for brain MRI classification using FCM and AE features.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 4.

Figure 4.

Augmented images were obtained through spatial and intensity augmentation.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 5.

Figure 5.

Brain MRI segmentation using FCM.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 6.

Figure 6.

AE architecture.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 7.

Figure 7.

Features extracted using AE.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 8.

Figure 8.

DNN classifier.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 9.

Figure 9.

Performance comparison.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Fig 10.

Figure 10.

The training and validation accuracy curves of the proposed method.

The International Journal of Fuzzy Logic and Intelligent Systems 2021; 21: 349-357https://doi.org/10.5391/IJFIS.2021.21.4.349

Share this article on :

Related articles in IJFIS

Most KeyWord