MRI-based brain tumor detection using convolutional deep learning methods and chosen machine learning techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Neuroscience

Background:

• Early detection of brain tumors is critical for effective treatment.
• Traditional biopsy methods require invasive surgery.
• Computational intelligence offers a non-invasive approach to tumor identification.

Purpose of the Study:

• To develop and compare deep learning and machine learning models for brain tumor classification.
• To accurately diagnose glioma, meningioma, and pituitary tumors using MRI data.
• To enable early and precise detection of brain tumors.

Main Methods:

• Utilized a dataset of 3264 MRI brain images.
• Applied preprocessing and augmentation techniques to the image data.
• Developed and trained a 2D Convolutional Neural Network (CNN) and a convolutional auto-encoder.
• Compared performance against six traditional machine learning techniques.

Main Results:

• The 2D CNN achieved a training accuracy of 96.47% and an average recall of 95%.
• The auto-encoder network achieved 95.63% training accuracy and 94% recall.
• Both deep learning models showed areas under the ROC curve of 0.99 or 1.
• The 2D CNN significantly outperformed traditional methods like MLP and KNN.

Conclusions:

• The proposed 2D CNN offers optimal accuracy and efficiency for brain tumor classification.
• The 2D CNN is less complex and suitable for clinical integration by radiologists.
• Deep learning methods show significant potential for improving early brain tumor detection.