Automated multi-class brain tumor types detection by extracting RICA based features and employing machine learning...

Area of Science:

• Medical Imaging
• Machine Learning in Oncology
• Computational Neuroscience

Background:

• Brain tumors are a leading global cause of cancer mortality.
• Early detection of brain tumors significantly increases patient survival rates.
• Accurate classification of tumor types (pituitary, meningioma, glioma) is crucial for effective treatment.

Purpose of the Study:

• To introduce a novel feature extraction method, reconstruction independent component analysis (RICA), for multi-class brain tumor detection.
• To evaluate the efficacy of machine learning classifiers, specifically Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA), in classifying brain tumor types using RICA features.
• To assess the diagnostic performance of the proposed RICA-based methodology for differentiating between pituitary, meningioma, and glioma.

Main Methods:

• Proposed a novel reconstruction independent component analysis (RICA) for feature extraction from brain tumor images.
• Employed Support Vector Machine (SVM) with quadratic and linear kernels, and Linear Discriminant Analysis (LDA) for classification.
• Utilized 10-fold cross-validation for robust training and testing data validation.

Main Results:

• Achieved high accuracy in multi-class brain tumor classification: pituitary (99.34%), meningioma (96.96%), and glioma (95.88%).
• Demonstrated excellent performance metrics, including sensitivity (97.78%), specificity (100%), and AUC (up to 0.9892 for pituitary).
• The RICA feature extraction method proved effective in distinguishing between different brain tumor types.

Conclusions:

• The proposed RICA feature-based methodology shows significant potential for improving the diagnostic efficiency of multi-class brain tumor detection.
• This approach can enhance prediction accuracy, contributing to better clinical outcomes.
• The study highlights the value of advanced feature extraction techniques combined with machine learning for brain tumor classification.