Detection of Glioma from Brain CT Scan images using Explainable AI based Ensemble Feature Extraction

This research explores the intervention of ML approaches to brain CT scans, with a focus on feature extraction and feature selection to improve diagnostic accuracy. Brain CT scans are essential for detecting and diagnosing a range of neurological conditions, such as tumors, strokes, and traumatic brain injuries. However, the complexity of these images requires sophisticated computational techniques to effectively identify the most relevant features. In this study, various feature extraction methods like Local Binary Patterns (LBP), Local Directional Patterns (LDP), Discrete Wavelet Transform (DWT), Hough Transform, Discrete Fourier Transform (DFT), and Binary Wavelet Transform (BWT) were used to systematically extract key features and many machine learning algorithms were used for image classification, where Random Forest and Catboost classfier both achieved the best accuracy of 0.9653. Simple machine learning approaches are computationally less expensive than Deep learning approaches. Moreover, here feature analysis have been performed to provide more insights regarding features using Principal Component Analysis (PCA) and feature importance graph. This approach not only improves diagnostic precision and speed but also provides more interpretable and clinically useful tools for healthcare professionals. By refining the process of feature selection, this study advances the creation of more trustworthy and precise neurology diagnostic techniques..