2023, Vol. 3, Issue 1, Part A
Advanced machine learning techniques for accurate brain tumor categorization from MRI image data
Author(s): Anubhuti Singh and Raghavendra Patidar
Abstract: This paper aims to classify brain malignancies using machine learning approaches utilizing the Figshare Brain Tumor Dataset, which contains T1-weighted contrast-enhanced MRI images that are categorized as glioma, meningioma, or pituitary. Establishing a trustworthy multi-class categorization system is critical for facilitating early diagnosis. The preprocessing stages included normalization, scaling, grayscale switching and noise reduction to guarantee a consistent dataset and enhance the model's training. Pituitary tumors were the most common type of tumor seen, indicating a class imbalance. To identify key handcrafted characteristics, we employed statistical descriptors such as mean intensity and kurtosis along with Gray-Level Concurrent Appearance Matrix (GLCM), Local Binary Patterns (LBP) and Histogram of inclined gradients (HOG). These features captured significant patterns depending on form and texture, which were useful for tumor differentiation. We optimized three machine learning algorithms—Random Forest, K-Nearest Neighbors (KNN) and Logistic Regression—using grid search and cross-validation. For this purpose, we evaluated the model using the F1-score, recall, accuracy and precision measures. In terms of accuracy (87.57%), F1-score (85.35%) and precision-recall balance (85.35%), the Random Forest model outperformed the other classifiers. Findings demonstrate the efficacy of learning in ensemble with meticulously extracted features for accurate brain tumor classification using Figshare dataset MRI data.
DOI: 10.22271/27084493.2023.v3.i1a.64
Pages: 78-85 | Views: 74 | Downloads: 36
Download Full Article: Click Here
How to cite this article:
Anubhuti Singh, Raghavendra Patidar. Advanced machine learning techniques for accurate brain tumor categorization from MRI image data. Int J Electron Microcircuits 2023;3(1):78-85. DOI: 10.22271/27084493.2023.v3.i1a.64
