Brain Tumor Segmentation from MRI Images Using Handcrafted Convolutional Neural Network.

Area of Science:

• * Medical Imaging Analysis
• * Artificial Intelligence in Medicine
• * Computational Neuroscience

Background:

• * Accurate brain tumor segmentation from MRI is crucial for clinical decision-making.
• * Existing methods using solely handcrafted or deep learning features have limitations.
• * Enhancing segmentation performance can improve patient diagnosis and treatment.

Purpose of the Study:

• * To develop and evaluate a novel hybrid approach for brain tumor segmentation.
• * To combine handcrafted features with deep learning features for improved accuracy.
• * To assess the performance of the hybrid method against traditional techniques.

Main Methods:

• * Extraction of handcrafted features (intensity, texture, shape) from MRI scans.
• * Development and training of a novel Convolutional Neural Network (CNN) architecture.
• * Fusion of handcrafted and CNN-extracted features into a new CNN pathway.
• * Performance evaluation using the Brain Tumor Segmentation (BraTS) dataset and metrics like Dice score, sensitivity, and specificity.

Main Results:

• * The proposed hybrid method significantly outperformed traditional handcrafted and individual CNN-based segmentation methods.
• * Integration of handcrafted features enhanced CNN performance, leading to a more robust solution.
• * Achieved high segmentation accuracy, Dice score, sensitivity, and specificity.

Conclusions:

• * The hybrid approach offers a superior method for brain tumor segmentation compared to existing techniques.
• * Combining handcrafted and deep learning features yields a more generalizable and accurate segmentation model.
• * The method shows significant promise for real-world clinical applications in neuro-oncology.