Preliminary research on abnormal brain detection by wavelet-energy and quantum- behaved PSO

Area of Science:

• Medical Imaging
• Machine Learning
• Biomedical Signal Processing

Background:

• Accurate and early detection of abnormal brains is crucial for effective treatment.
• Wavelet-energy (WE) has shown promise as a feature descriptor in various applications.
• Existing methods for automated brain abnormality detection require improvement in performance.

Purpose of the Study:

• To propose a novel wavelet-energy (WE) based approach for automated abnormal brain detection.
• To optimize the kernel support vector machine (KSVM) classifier using quantum-behaved particle swarm optimization (QPSO).
• To evaluate the performance of the proposed WE + QPSO-KSVM method against established techniques.

Main Methods:

• Feature extraction using wavelet-energy (WE).
• Classification using kernel support vector machine (KSVM).
• Optimization of KSVM weights with quantum-behaved particle swarm optimization (QPSO).
• Performance evaluation using 5x5-fold cross-validation.

Main Results:

• The proposed WE + QPSO-KSVM method significantly outperformed "DWT + PCA + BP-NN", "DWT + PCA + RBF-NN", "DWT + PCA + PSO-KSVM", "WE + BPNN", "WE + KSVM", and "DWT + PCA + GA-KSVM".
• The WE + QPSO-KSVM achieved superior sensitivity, specificity, and accuracy.
• Preliminary results indicate excellent performance of the developed approach.

Conclusions:

• The WE + QPSO-KSVM presents a novel and effective method for automated abnormal brain detection.
• This approach offers a significant advancement in the field of medical image analysis for neurological disorders.
• The study highlights the potential of combining advanced feature descriptors with optimized machine learning classifiers.