An Intelligent Diagnosis Method of Brain MRI Tumor Segmentation Using Deep Convolutional Neural Network and SVM...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Traditional brain segmentation methods using image processing and machine learning show limitations.
• Deep learning, particularly convolutional neural networks (CNNs), demonstrates promise for brain tumor segmentation.
• Existing deep CNN models face challenges with numerous parameters and information loss during encoding/decoding.

Purpose of the Study:

• To propose a novel deep convolutional neural network fusion support vector machine (DCNN-F-SVM) algorithm for enhanced brain tumor segmentation.
• To address the limitations of existing deep CNN models in brain tumor segmentation.
• To evaluate the performance of the proposed DCNN-F-SVM model against established methods.

Main Methods:

• A three-stage approach integrating a deep convolutional neural network (CNN) with a support vector machine (SVM).
• Stage 1: CNN training to map image space to tumor marker space.
• Stage 2 & 3: Sequential integration of CNN predictions and test images into an SVM classifier to train a deep classifier.

Main Results:

• The proposed DCNN-F-SVM model was evaluated on the BraTS dataset and a self-made dataset for brain tumor segmentation.
• Segmentation results demonstrated significantly improved performance of the DCNN-F-SVM model.
• The DCNN-F-SVM outperformed both standalone deep convolutional neural networks and integrated SVM classifiers.

Conclusions:

• The DCNN-F-SVM algorithm represents a significant advancement in brain tumor segmentation techniques.
• The fusion approach effectively overcomes the limitations of individual deep learning and machine learning models.
• This method offers a more robust and accurate solution for clinical brain tumor segmentation applications.