Brain Pathology Classification of MR Images: Comparison
View Latest Version
Please note this is a comparison between Version 2 by Catherine Yang and Version 1 by Muhammad Syafrudin.

A brain tumor is essentially a collection of aberrant tissues, so it is crucial to classify tumors of the brain using MRI before beginning therapy. Tumor segmentation and classification from brain MRI scans using machine learning techniques are widely recognized as challenging and important tasks. The potential applications of machine learning in diagnostics, preoperative planning, and postoperative evaluations are substantial. Accurate determination of the tumor’s location on a brain MRI is of paramount importance. 

  • machine learning
  • tumor segmentation
  • classification

1. Introduction

The brain holds paramount significance as the body’s central controller and regulator, making it the most vital organ. Tumors are aberrant masses of tissue that arise due to uncontrolled cell division. It is still unknown what triggers brain cancer. Standard imaging methods often face significant challenges in accurately detecting aberrant brain anatomy in humans. MRI methods aid in elucidating the complex neural structure of the human brain [1]. To simplify matters and boost segmentation and measurement performance when tumor size, position, and shape change, we must first pay attention to noise reduction, color visualization of the brain tumor region, segmentation, size measurements, and classification [2]. After the segments have been extracted, morphological filtering can be used to eliminate any remaining noise. The potential use of high-precision segmentation in diagnosing cancer in tumors has also been proposed.
Image importance may be determined with the use of techniques like principal component analysis (PCA) and deep wavelet transforms (D-DWTs) in two dimensions (PCA). The classification was achieved using a Feed-forward Neural Network (FNN) and a K-Nearest Neighbor (KNN) method. Using features fed into a least squares support vector machine classifier, Das et al. [3] created a Ripple Transform (RT) model. Fluid vector and T1 weighted pictures are two potential approaches for tumor identification [4]. The tumor location was determined using diffusion tensor imaging in combination with diffusion coefficients [5]. Due to the exponential rise in the number of available traits, it has been difficult for researchers studying brain tumors to identify and eliminate the most conspicuous feature. The selection of informative training and testing samples is also difficult [6,7][6][7]. Amin et al. [8] developed a novel method of MR brain categorization. Gaussian-filtered images of the brain were processed further to remove any remaining noise. After that, segmentation processing, including the extraction of embedding, cyclic, contrast, and block appearance features, and a cross-validation strategy were used to accomplish the classification. The fuzzy clustering membership from the original picture is included in the Markov random field’s function, as described by [9,10][9][10]. This method has shown to be successful due to the use of a hybrid strategy and segregated supporting data.
The maximum likelihood and minimum distance classification techniques are two well-liked supervised approaches. It is also common practice to employ support vector machines (SVMs). Regarding supervised classification, support vector machines are the gold standard. Nonetheless, SVMs may be used unsupervised as well [9]. Researchers have suggested using an improved support vector machine (ISVM) classifier to classify brain cancers [11,12][11][12]. The proposed approach uses a dataset of images processed through the K-means segmentation technique to identify abnormal cells in MRI scans as cancer.

2. Brain Pathology Classification of MR Images Using Machine Learning Techniques

Brain pathology classification using MRI has gained significant attention recently due to its potential in early disease detection. Machine learning techniques have been widely explored to automate and improve the accuracy of brain pathology classification.  A Hybrid Ensemble Model classifier has been suggested by Garg and his colleagues [15][13] to classify brain tumors. Their model includes Random Forest (RF), K-Nearest Neighbor, and Decision Tree (DT) (KNNRF-DT) based on the Majority Voting Method with an accuracy of about 97.305%. However, their model lacks expressiveness in Decision Trees, so it may need to be more expressive to capture complex relationships in the data. In their research [16][14], Kesav and his team introduced an alternative model based on the RCNN technique. Their proposed model was evaluated using two publicly available datasets with the two-channel CNN approach. The primary objective was to reduce the computational time of the conventional RCNN architecture by employing a simpler framework while creating a system for brain tumor analysis. Remarkably, their study achieved an accuracy of 97.83%. Nevertheless, effective training of CNNs requires a substantial amount of labeled data, which can be difficult to obtain in the context of brain tumor classification. Privacy concerns, limited access to medical data, and the necessity for expert annotations all contribute to the complexity of acquiring a large and diverse dataset. Deep Hybrid Boosted uses a two-phase deep-learning-based framework suggested by Khan and his colleagues [17][15] to detect and categorize brain tumors in magnetic resonance images (MRIs); they reached 95% accuracy. Although they did not reach the best accuracy in the field, the Complexity of Hybrid feature fusion methods can be difficult to implement and require careful design and tuning. Integrating different types of features and deciding on the fusion strategy can be challenging and may lead to increased model complexity. Alanazi and his colleagues [18][16] developed a transfer-learned model with various layers of isolated convolutional neural network (CNN). According to them, the models are built from scratch to check their performances for brain MRI images. The model has also been tested using the brain MRI images of another machine to validate its modification. The accuracy of the model is 95%. Effective transfer learning often requires a substantial amount of labeled data for fine-tuning. In medical applications, acquiring a large and diverse dataset can be particularly arduous, especially when dealing with rare conditions. The need for such data may limit the overall efficacy of the transfer learning approach in these scenarios.

3. Future Work

Manually analyzing brain scans is laborious and is becoming more irrelevant as ourthe understanding of the brain expands. Alternatively, automated segmentation and categorization make neurologists’ jobs easier by speeding up the decision-making process. Recent advances in the segmentation and prediction of brain tumor MRI images have been made possible via the use of deep learning algorithms. Despite this, MRI is still a complex topic where additional study is needed. Medical practitioners benefit substantially from segmentation and classification procedures since it expedites data analysis and provides a second opinion based on automated findings. Image processing methods pioneered by FCMT, such as contrast enhancement, are used in tumor segmentation. Features are extracted using LBP and HOG. All of these characteristics come together to form an ensemble. In order to train the neural network, this Ensemble Features set is subjected to convolutional training. The accuracy of an SVM merged with HOG and LPB features is shown to be 99.9%, whereas that of a retrained CNN model is 98%. Less than 2% of errors are discovered. 

References

  1. Biratu, E.S.; Schwenker, F.; Ayano, Y.M.; Debelee, T.G. A survey of brain tumor segmentation and classification algorithms. J. Imaging 2021, 7, 179.
  2. El-Dahshan, E.-S.A.; Hosny, T.; Salem, A.-B.M. Hybrid intelligent techniques for MRI brain images classification. Digit. Signal Process. 2010, 20, 433–441.
  3. Das, S.; Chowdhury, M.; Kundu, M.K. Brain MR image classification using multiscale geometric analysis of ripplet. Prog. Electromagn. Res. 2013, 137, 1–17.
  4. Hamad, Y.A.; Seno, M.E.; Al-Kubaisi, M.; Safonova, A.N. Segmentation and measurement of lung pathological changes for COVID-19 diagnosis based on computed tomography. Period. Eng. Nat. Sci. (PEN) 2021, 9, 29–41.
  5. Al-Okaili, R.N.; Krejza, J.; Woo, J.H.; Wolf, R.L.; O’Rourke, D.M.; Judy, K.D.; Poptani, H.; Melhem, E.R. Intraaxial brain masses: MR imaging-based diagnostic strategy—Initial experience. Radiology 2007, 243, 539–550.
  6. Jiang, J.; Wu, Y.; Huang, M.; Yang, W.; Chen, W.; Feng, Q. 3D brain tumor segmentation in multimodal MR images based on learning population- and patient-specific feature sets. Comput. Med. Imaging Graph. 2013, 37, 512–521.
  7. Ortiz, A.; Górriz, J.M.; Ramírez, J.; Salas-Gonzalez, D. Improving MRI segmentation with probabilistic GHSOM and multiobjective optimization. Neurocomputing 2013, 114, 118–131.
  8. Zotin, A.; Hamad, Y.; Simonov, K.; Kurako, M.; Kents, A. Processing of CT Lung Images as a Part of Radiomics. In International Conference on Intelligent Decision Technologies; Springer: Singapore, 2020; pp. 243–252.
  9. Chen, M.; Yan, Q.; Qin, M. A segmentation of brain MRI images utilizing intensity and contextual information by Markov random field. Comput. Assist. Surg. 2017, 22, 200–211.
  10. Raja, P.S.; Rani, A.V. Brain tumor classification using a hybrid deep autoencoder with Bayesian fuzzy clustering-based segmentation approach. Biocybern. Biomed. Eng. 2020, 40, 440–453.
  11. Hamad, Y.A.; Simonov, K.V.; Naeem, M.B. Detection of brain tumor in MRI images, using a combination of fuzzy C-means and thresholding. Int. J. Adv. Pervasive Ubiquitous Comput. (IJAPUC) 2019, 11, 45–60.
  12. Deepak, S.; Ameer, P. Brain tumor classification using deep CNN features via transfer learning. Comput. Biol. Med. 2019, 111, 103345.
  13. Garg, G.; Garg, R. Brain tumor detection and classification based on hybrid ensemble classifier. arXiv 2021, arXiv:2101.00216.
  14. Kesav, N.; Jibukumar, M. Efficient and low complex architecture for detection and classification of brain tumor using RCNN with two channel CNN. J. King Saud Univ.-Comput. Inf. Sci. 2021, 34, 6229–6242.
  15. Khan, M.F.; Khatri, P.; Lenka, S.; Anuhya, D.; Sanyal, A. Detection of Brain Tumor from the MRI Images using Deep Hybrid Boosted based on Ensemble Techniques. In Proceedings of the 2022 3rd International Conference on Smart Electronics and Communication (ICOSEC), Ttichy, India, 20–22 October 2022; pp. 1464–1467.
  16. Alanazi, M.F.; Ali, M.U.; Hussain, S.J.; Zafar, A.; Mohatram, M.; Irfan, M.; AlRuwaili, R.; Alruwaili, M.; Ali, N.H.; Albarrak, A.M. Brain tumor/mass classification framework using magnetic-resonance-imaging-based isolated and developed transfer deep-learning model. Sensors 2022, 22, 372.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/)
Video Production Service
Feedback