Segmentation of Brain Tumor Using a 3D Generative Adversarial Network.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuro-oncology

Background:

• Brain tumor detection in MRI scans is challenging due to subtle visual cues and limited labeled datasets.
• Manual segmentation of medical images is time-consuming and labor-intensive, hindering large-scale analysis.

Purpose of the Study:

• To propose an accurate segmentation method for 3D MRI brain images to precisely identify tumor locations.
• To leverage Generative Adversarial Networks (GANs) for improved brain tumor segmentation.

Main Methods:

• Utilized a 3D Generative Adversarial Network (GAN) as a classification network for detecting and segmenting brain tumors in MRI data.
• Implemented mutual training within the GAN framework to refine segmentation accuracy against labeled data.
• Compared two experimental models using the BraTS 2021 dataset of 3D brain MRI images.

Main Results:

• The 3D GAN model demonstrated dense connectivity, enabling rapid network convergence and enhanced information extraction.
• Mutual training in the GAN approach improved segmentation results, aligning them more closely with labeled data.
• The proposed method showed potential for accurate tumor segmentation in 3D MRI scans.

Conclusions:

• The developed 3D GAN segmentation method offers a promising approach for accurate brain tumor identification in medical imaging.
• This technique addresses limitations of manual segmentation and small datasets, potentially improving diagnostic capabilities.
• Further validation and comparison with existing models were conducted using the BraTS 2021 dataset.