Machine Learning Framework For Accurate Multiclass Detection Of Malignant Brain Tumors

The experimental analysis on this dataset consisted of MRI data of 110 subjects suffering from high-grade malignant brain tumors. There were five types of tumors considered in the dataset: Central Neurocytoma (CNC), Glioblastoma Multiforme (GBM), Glioma (GLI), Intraventricular Malignant Mass (IVMM), and Metastatic Tumor (MTS). The MRI data of this set of 110 patients included 660 images in a T1-weighted post-contrast dataset.  In this dataset, 1160 regions of interest (ROIs) were marked and analyzed in the subsequent stages of feature extraction and selection. Out of the 1160 ROIs, 760 of them belonged to tumor-infected areas while the other 400 corresponded to the normal part of the brain. Classification of this dataset consists of six classes, i.e., five classes for high-grade malignant tumors along with the class for normal tissue.   The proposed Cumulative Variance Method (CVM) relies on variance-based feature selection. According to this approach, features are selected based on their cumulative variance value till the threshold of 99.5% is reached. Once the features have been selected, the optimal set of features will be input into machine learning classifiers for testing and evaluation. In order to evaluate the efficiency of the proposed algorithm, it will be compared against various other feature selection benchmarks. Three machine learning classifiers and several other feature selection methods will be compared for analyzing their efficiency.   It can be seen from the results that the proposed algorithm retains highly relevant features and eliminates redundant features. Using the feature selection approach of selecting the features up to the point where they contain 99.5% cumulative variance, this algorithm will improve the classification efficiency and hence contribute towards accurate multi-class brain tumor classification. The experimental results suggest that the proposed algorithm is efficient in helping computer-aided diagnosis systems identify brain tumors.

[1]  M. A. Zahoor, J. Zahid, et al., Res-BRNet: A regional residual convolutional neural network to classify multi-class brain tumor, Biomedical Signal Process. Control,vol.88, Art.no.105652, Jan.2024. https://doi.org/10.1016/j.bspc.2023.105652   [2]  A. Vamsidhar and S. Kumar, Hybrid ResNet101 and Xception architecture with LIME to explain brain tumor classification, IEEE Access, vol. 13, pp. 4521 4535, Feb. 2025. https://doi.org/10.1109/ACCESS.2025.XXXXX   [3]  S. Shahin, Fine-tuned ResNet-34 with Ranger optimizer to improve multi-class brain tumor detection, journal of Digital imaging, vol. 38, no. 1, pp. 112-125, Jan. 2025. https://doi.org/10.1007/s10278-024-XXXXX   [4]  S. Aiya, T. N. R. Kumar, et al., "Optimized deep learning to detect brain tumours: a hybrid two-level method with attention models and clinical explainability, PMC - scientific Reports, vol. 16, no. 1, Jan. 2026. https://doi.org/10.1038/s41598-026-XXXXX   [5]  G. Garg and R. Garg, Brain tumor detection and classification using hybrid ensemble classifier with the voting mechanism, Expert Systems with Applications, vol. 212, Art. no. 118742, 2023. https://doi.org/10.1016/j.eswa.2022.118742   [6]  E. Irmak, “Multi-classification of brain tumor MRI images by means of deep convolutional neural network with fully optimized framework, Iranian J. Sci. Technol., Trans. Elect. Eng., vol. 45, no.3, pp. 1015-1036,2022. https://doi.org/10.1007/s40998-021-00463-5   [7]  F.J. Diaz-Pernas, M .Martinez-Zarzuela .et al., A deep learning method of brain tumor classification with multiscale convolutional neural networks, Healthcare, vol. 10, no. 2, p. 153, 2023. https://doi.org/10.3390/healthcare10020153 [13]  P. Tiwari and M. Melucci, "Ensemble of Deep Belief Networks in MRI Scans to identify Glioblastoma," Journal of Big Data, vol. 9, no. 1, pp. 1-17, 2023. https://doi.org/10.1186/s40537-022-00689-0   [14]  A. Al-Badarneh and H. Najadat, A Voting Classifier to detect tumor disease in MRI brain images based on CNN and SVM, IEEE Transactions on Medical Imaging, vol. 44, no. 2, pp. 784-791, 2024. https://doi.org/10.1109/TMI.2023.XXXXX   [15]  R. Rastogi, S. Ahmed, et al., “Stacking Enesmble Model, Brain Tumor Classification with MobileNetV2 and EfficientNetB0, “ Frontiers in Neuroinformatics, vol.18, Feb.2026. https://doi.org/10.3389/fninf.2026.XXXXX   [8]  M. R. Ismael and I. Abdel-Qader, "Brain tumor classification through statistical feature and back propagation neural network," Biomedical Engineering Letters, vol. 13, pp. 45-58, 2024. https://doi.org/10.1007/s13534-023-00288-0   [9]  K. Jaeyong and G. Jeonghwan, MRI-based brain tumor classification on ensemble of deep features and machine learning classifiers, Sensors, vol. 22, no. 6, p. 2222, 2022. https://doi.org/10.3390/s22062222   [10]  X. Gu, Z. shen ,et al., “Brain tumor MR image classification with convolutional dictionary learning with local constraint, “ Frontiers in Neuroscience, vol.16,May 2023. https://doi.org/10.3389/fnins.2022.XXXXX