Feature extraction from MRI ADC images for brain tumor classification using machine learning techniques

Area of Science:

• Radiology and Medical Imaging
• Machine Learning in Healthcare
• Oncology

Background:

• Diffusion-weighted (DW) imaging is a standard MRI technique for brain examinations.
• Apparent Diffusion Coefficient (ADC) images offer rich data for tumor analysis.
• Distinguishing malignant from benign brain tumors is crucial for treatment planning.

Purpose of the Study:

• To extract demographic and texture features from brain tumor MRI ADC images.
• To identify feature distribution patterns.
• To apply Machine Learning (ML) for differentiating malignant and benign brain tumors.

Main Methods:

• A prospective study analyzed 1599 MRI brain ADC images from 195 patients.
• Demographic, pixel value, and Grey Level Co-occurrence Matrix (GLCM) features were extracted.
• ANOVA f-test was used for feature selection, followed by ML classification (Random Forest).

Main Results:

• Skewness and GLCM homogeneity were excluded due to low ANOVA f-test scores.
• The Random Forest classifier achieved the highest accuracy.
• The final ML model accurately predicted tumor malignancy with 90.41% accuracy.

Conclusions:

• Selected features (excluding skewness and GLCM homogeneity) are effective for brain tumor differentiation.
• A high-performance ML model can assist in pre-biopsy diagnostic decision-making.
• This approach may reduce the need for invasive diagnostic procedures.