Elsevier

Biocybernetics and Biomedical Engineering

Volume 40, Issue 3, July–September 2020, Pages 1022-1035
Biocybernetics and Biomedical Engineering

Original Research Article
A novel privacy-supporting 2-class classification technique for brain MRI images

https://doi.org/10.1016/j.bbe.2020.05.005Get rights and content

Abstract

Developing automated Computer Aided Diagnosis (CAD) framework for assisting radiologists in a fast and effective classification of brain Magnetic Resonance (MR) images is of great importance, given plausible usage of Electronic Health Records (EHR) in healthcare system. This work aims at proposing two novel privacy supporting classifiers for automatic segregation of brain MR images. To ensure privacy, our article employs a spatial steganographic approach to hide patients sensitive health information in brain images itself. Proposed methods employ Discrete Wavelet Transform (DWT) for extracting relevant features from original and stego images. Subsequently, Symmetrical Uncertainty Ranking (SUR) and Probabilistic Principal Components Analysis (PPCA) are used to obtain a reduced feature vector for Support Vector Machine (SVM) and Filtered Classifier (FC) respectively. The experiments are carried out on two benchmark datasets DS-75 and DS-160 collected from Harvard Medical School website and one larger input pool of self-collected dataset NITR-DHH. To validate this work, the proposed schemes are experimented on both original and stego brain MR images and are compared against eight state-of-the-art classification techniques with respect to six standard parameters. The results reveal that the proposed techniques are robust and scalable with respect to the size of the datasets. Moreover, it is concluded that applying steganographic algorithm on brain MR images yield equally satisfactory classification rate.

Introduction

The growth rate of brain diseases among children and adults are increasing exponentially day-by-day. Hence quick and accurate detection of brain disease can be very helpful to maintain a healthy lifestyle. The Electronic Health Records (EHR) of the patients automate the access to important healthcare information and can improve patient care by streamlining the clinical process. A varieties of medical imaging techniques, viz., Computed Tomography (CT), Positron Emission Tomography (PET), and Magnetic Resonance Imaging (MRI) produce visual depiction of the internal substances of brain that helps in proper medical diagnosis. MRI is more significant and suitable than other imaging tools due to its superior contrast, non-invasive, and non-radiation properties [1]. In addition to that, different soft tissues, white, and gray matters of the brain are more clearly visible via MRI vis-a-vis other brain imaging techniques.

Manual segmentation becomes time consuming and tedious for large pool of brain MR images [2]. This necessitates the design of Computer Aided Diagnosis (CAD) systems for assisting radiologists [3] to accurately classify brain MR images (symmetry-asymmetry detection). At the same time, CAD can reduce the burden of increasing workload for the medical personnel, also supports in improving the classification accuracy by implementing some machine learning techniques [4]. CAD is a diagnostic system which integrates various image processing techniques like feature extraction, selection, and classification of images for the pathological brain detection. Moreover, it is a tough task to devise an accurate and efficient CAD system regardless of the size and the quality of the brain MR images. CAD systems are primarily designed for improving maximum possible accuracy while classifying brain MR images.

Various studies on automated CAD system categorize it into two basic types, viz., a. CAD systems related to classify brain MR images, b. CAD systems to differentiate between benign and malign diseases. The main objective behind this is to design a fast and reliable CAD system to enhance diagnostic accuracy of radiologists and at the same time reducing the burden of increasing workload. To improve the accuracy of the CAD systems, correct amalgamation of feature extractor, selector, and classifier plays a crucial role. Moreover, all imaging modalities contain patients’ private information. Hence, privacy issues, in addition to classification accuracy are also critical and an open challenge when a healthcare provider plans to deploy an online medical record management system. In Internet of Medical Things (IoMT), privacy and confidentiality of patient's sensitive information must be preserved and only accessible to authorized parties [5]. Therefore, it is essential to develop a relatable CAD system to achieve clinically acceptable classification accuracy with reduced feature count along with addressing such privacy issues and challenges. Motivated by the above prime requirements, we devise automated CAD systems (with a novel combination of extractor, selector, and classifier) to assist the radiologists for quicker and easier diagnosis of brain diseases without compromising privacy of patients’ sensitive information.

The article is organized as follows. Section 2 provides some of the existing brain MR image classification methods. Section 3 describes two new proposed frameworks for implementing privacy supporting two class classifiers. Section 4 demonstrates the experimental analysis and comparative result of both proposed frameworks on three different datasets. Finally, Section 5, presents the importance of our proposed techniques with an outline of possible future augmentations to this exploration.

Section snippets

Related work

The state-of-the-art research challenges related to CAD and brain MR image classifier systems have been reviewed and presented in this section. In the past years, a significant amount of work has been done in segregating normal brain from pathological brain. In almost all Pathological Brain Detection Systems (PBDSs), MRI has been extensively used as an imaging technique, as it provides ample amount of information about different soft tissues of the brain. Wavelet Transform is the most

Proposed methodologies

Different researchers have tried different combinations of feature extraction, selection, and classification mechanisms to find satisfactory accuracy in brain MR image classification. With a quest to have equivalent classification, we have used DWT as feature extractor, SUR and Probabilistic Principal Components Analysis (PPCA) as feature selectors, and SVM and Filtered Classifier (FC) as 2-class classifiers. Further, we have employed handcrafted features since these feature vectors are best

Datasets used

We have reviewed four brain MR image datasets from three different sources. The details on all these datasets are described in Table 1.

The 3D dataset Harvard GSP [26] sourced from Harvard dataverse contains 1500 brain genomics of T1-weighted MRI. These datasets are primarily used for medical image registration [25], and have 256 × 256 × 256 in-plane resolutions. So, this dataset is not suitable for our experiments. To initially verify the performance of brain abnormality classification, we have

Conclusion

In this work, we propose two new privacy supporting binary classifier systems for automated detection of pathological brain in MR images. The novelty of our work lies in guaranteeing the privacy of patients EHR without much affecting the classification rate. We adopt LSB substitution steganographic method and visually hide private information of a patient by watermarking it on the original MR images itself prior to classification. We focus on achieving better classification rate towards

Acknowledgment

This research is partially supported by the following project titled "Deep learning applications for computer vision task" funded by NITROAA with support of Lenovo P920 workstation and NVIDIA Corporation with support of NVIDIA Titan V GPU.

References (64)

  • D.R. Nayak et al.

    A hybrid regularized extreme learning machine for automated detection of pathological brain

    Biocybern Biomed Eng

    (2019)
  • A. Narayanan et al.

    Multi-channeled MR brain image segmentation: a novel double optimization approach combined with clustering technique for tumor identification and tissue segmentation

    Biocybern Biomed Eng

    (2019)
  • M. Talo et al.

    Application of deep transfer learning for automated brain abnormality classification using MR images

    Cognit Syst Res

    (2019)
  • M. Sajjad et al.

    Multi-grade brain tumor classification using deep CNN with extensive data augmentation

    J Comput Sci

    (2019)
  • F. Ozyurt et al.

    Brain tumor detection based on convolutional neural network with neutrosophic expert maximum fuzzy sure entropy

    Measurement

    (2019)
  • H. Tekchandani et al.

    Mediastinal lymph node malignancy detection in computed tomography images using fully convolutional network

    Biocybern Biomed Eng

    (2020)
  • J.R. Reichenbach

    The impact of deep learning

    Z Med Phys

    (2019)
  • A. Subudhi et al.

    Automated segmentation and classification of brain stroke using expectation-maximization and random forest classifier

    Biocybern Biomed Eng

    (2020)
  • S. Deepak et al.

    Brain tumor classification using deep CNN features via transfer learning

    Comput Biol Med

    (2019)
  • S. Lu et al.

    Pathological brain detection based on AlexNet and transfer learning

    J Comput Sci

    (2019)
  • A. Cheddad et al.

    Digital image steganography: survey and analysis of current methods

    Signal Process

    (2010)
  • C.K. Chan et al.

    Hiding data in images by simple LSB substitution

    Pattern Recognit

    (2004)
  • L. Nanni et al.

    Handcrafted vs. non-handcrafted features for computer vision classification

    Pattern Recognit

    (2017)
  • G. Chandrashekar et al.

    A survey on feature selection methods

    Comput Electr Eng

    (2014)
  • S. Devi et al.

    Hiding medical information in brain MR images without affecting accuracy of classifying pathological brain

    Future Gen Comput Syst

    (2019)
  • S. Chapaneri et al.

    Evaluation of music features for PUK kernel based genre classification

  • M. Talo et al.

    Convolutional neural networks for multi-class brain disease detection using MRI images

    Comput Med Imaging Graph

    (2019)
  • V. Cheplygina

    Cats or CAT scans: transfer learning from natural or medical image source data sets?

    Curr Opin Biomed Eng

    (2019)
  • Z. Ji et al.

    Robust spatially constrained fuzzy c-means algorithm for brain MR image segmentation

    Pattern Recognit

    (2014)
  • C. Westbrook

    Handbook of MRI technique

    (2008)
  • V. Barra et al.

    Automatic segmentation of subcortical brain structure in MR images using information fusion

    IEEE Trans Med Imaging

    (2001)
  • R. Duda et al.

    Pattern classification. Pattern classification

    (2001)
  • Cited by (9)

    • Effect of situational and instrumental distortions on the classification of brain MR images

      2023, Biomedical Signal Processing and Control
      Citation Excerpt :

      Brain Magnetic Resonance (MR) images clearly specify the structural details of different inner parts of the brain, thereby assisting the radiologists in better and accurate clinical diagnosis [1,2] and prognosis. However, manual categorization of “normal” and “diseased” brain becomes tedious and time consuming for larger input pool of brain MR images [3,4]. This requires the development of automated Computer Aided Diagnosis (CAD) systems for aiding radiologists [5,6] in precisely classifying brain MR images.

    • Assessment of CT for the categorization of hemorrhagic stroke (HS) and cerebral amyloid angiopathy hemorrhage (CAAH): A review

      2022, Biocybernetics and Biomedical Engineering
      Citation Excerpt :

      In recent years, we have observed increasing interest among researchers in developing machine learning (ML) and deep learning (DL) based methods for CT image classification and hematoma segmentation [41-45]. In addition, several recent large-scale studies have focused on detecting ICH in the acute setting, aiming to optimize the clinical workflow and reduce the time to diagnosis [46-48], and mainly targeting brain imaging [48-58]. Furthermore, ML methods have drawn much attention among prediction models due to their high predictive power and reliable performance.

    • Brain MR Image Classification Using Superpixel-Based Deep Transfer Learning

      2024, IEEE Journal of Biomedical and Health Informatics
    View all citing articles on Scopus
    View full text