Brain Tumor Classification Using Meta-Heuristic Optimized Convolutional Neural Networks.

Area of Science:

• Medical Image Processing
• Artificial Intelligence in Medicine
• Neurology

Background:

• Early brain tumor diagnosis is crucial for patient survival.
• Existing automatic systems struggle with precise tumor region identification and edge detail detection.
• Computational complexity is a concern in current medical image analysis systems.

Purpose of the Study:

• To develop a more efficient system for brain tumor classification using medical image processing.
• To improve the accuracy of tumor region identification and preserve hidden edge details.
• To reduce computational complexity in automated tumor recognition.

Main Methods:

• Pre-processing of brain magnetic resonance (MR) images to remove noise.
• Utilizing a candidate region method with line segments to identify tumor boundaries and preserve edge details.
• Applying a convolutional neural network (CNN) optimized with the Harris Hawks algorithm (HHOCNN) for feature extraction and classification.
• Performance evaluation using metrics like pixel accuracy, error rate, accuracy, specificity, and sensitivity.

Main Results:

• The proposed HHOCNN system effectively minimizes false tumor recognition by eliminating noisy pixels.
• The candidate region method successfully identifies tumor regions while retaining hidden edge details.
• The HHOCNN achieved a high tumor recognition accuracy of 98% on the Kaggle dataset.
• The Harris Hawks optimization significantly reduced misclassification error rates.

Conclusions:

• The HHOCNN offers an efficient and accurate solution for brain tumor classification from MR images.
• This method enhances the identification of exact tumor regions and subtle edge details.
• The developed system demonstrates potential for improving early diagnosis and patient outcomes in neuro-oncology.