An Improved Brain MRI Classification Methodology Based on Statistical Features and Machine Learning Algorithms.

Area of Science:

• Medical Imaging
• Machine Learning
• Image Processing

Background:

• Magnetic Resonance Imaging (MRI) images are susceptible to salt-and-pepper noise.
• Image quality is crucial for accurate analysis and diagnosis.
• Existing methods may lack efficiency or accuracy in feature extraction and classification.

Purpose of the Study:

• To propose a novel methodology for image analysis using statistical features and machine learning algorithms.
• To enhance the quality of MRI images through preprocessing techniques.
• To evaluate the performance of different machine learning algorithms for image classification.

Main Methods:

• Preprocessing: Median filter for noise reduction and histogram equalization for image enhancement. Grayscale to RGB conversion.
• Feature Extraction: Calculation of statistical measures (mean, variance, skewness, kurtosis, entropy, energy, contrast, homogeneity, correlation) for each RGB channel, yielding 27 features per image.
• Classification: Application of Artificial Neural Network, K-Nearest Neighbors, Naïve Bayes, and Decision Tree algorithms.

Main Results:

• The Decision Tree algorithm outperformed other machine learning algorithms in the classification stage.
• The proposed methodology demonstrated superior accuracy and simplicity compared to existing well-known algorithms.
• Extracted statistical features effectively represented image characteristics for classification.

Conclusions:

• The proposed methodology, utilizing statistical features and a Decision Tree classifier, offers an effective approach for image analysis.
• The method provides enhanced image quality and accurate classification, outperforming existing techniques.
• This approach holds potential for improving diagnostic accuracy in medical imaging applications.