EGDP based feature extraction and deep convolutional belief network for brain tumor detection using MRI image.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Brain tumour (BT) detection from MRI is crucial for timely diagnosis and treatment.
• Existing methods may face challenges with noise, artefacts, and accurate segmentation.
• Deep learning offers potential for improved automated BT detection.

Purpose of the Study:

• To develop and evaluate a novel deep learning framework for MRI-based brain tumour detection.
• To enhance the accuracy and reliability of automated brain tumour identification.
• To introduce a new network architecture combining deep convolutional neural networks and deep belief networks.

Main Methods:

• Utilized a dataset of brain MRI images.
• Applied median filtering for noise and artefact removal.
• Employed RP-Net for brain tumour region segmentation.
• Implemented image augmentation techniques (rotation, flipping, shifting, colour augmentation) to prevent overfitting.
• Extracted features using Gray-Level Co-occurrence Matrix (GLCM) and a novel Entropy-based Gray Difference Probability (EGDP).
• Developed and applied a deep convolutional belief network (DCvB-Net) for final brain tumour detection.

Main Results:

• The proposed DCvB-Net achieved a true negative rate of 93%, accuracy of 92.3%, and a true positive rate of 93.1%.
• The framework demonstrated effective noise reduction and accurate tumour segmentation.
• Feature extraction methods, including EGDP, contributed to robust detection.

Conclusions:

• The novel deep learning framework, DCvB-Net, shows significant promise for accurate and reliable MRI-based brain tumour detection.
• The integrated approach of preprocessing, segmentation, feature extraction, and a hybrid deep network is effective.
• This research contributes a valuable tool for radiologists and oncologists in diagnosing brain tumours.