Characterizing Brain Tumor Regions Using Texture Analysis in Magnetic Resonance Imaging

Area of Science:

• Radiology
• Medical Imaging
• Machine Learning

Background:

• Brain tumors require accurate and early diagnosis.
• Magnetic resonance imaging (MRI) offers detailed anatomical information.
• Texture analysis can quantify image characteristics for diagnostic support.

Purpose of the Study:

• To extract texture features from brain MRI scans.
• To develop a classification model for early brain tumor diagnosis.
• To evaluate the efficacy of different machine learning classifiers.

Main Methods:

• Texture features were extracted from MRI scans of 40 patients (meningioma/glioma).
• Statistical tests identified significant differences between tumor and healthy regions.
• Random forests (RFs) selected top features; SVM, RF, and back propagation (BP) models were trained.

Main Results:

• Sixteen texture features showed significant differences between tumor and healthy areas.
• Top features included standard deviation, variance, and moments.
• The SVM classifier achieved the highest performance (94.04% sensitivity, 92.3% specificity, 0.932 AUC).

Conclusions:

• Texture analysis combined with SVM provides a fast and accurate method for differentiating brain tumors.
• This approach supports clinical applications for early tumor detection.
• High diagnostic performance indicates potential for integration into clinical workflows.