RAUM-GANs: a multi-layer GAN-enhanced framework for accurate multiple sclerosis lesion segmentation in MRI

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuroscience

Background:

• Multiple sclerosis (MS) presents challenges for MRI-based lesion segmentation due to image noise, missing data, and limited high-quality labeled datasets.
• Accurate segmentation is crucial for diagnosing and monitoring MS progression.

Purpose of the Study:

• To introduce RAUM-GANs, a multi-layer deep learning framework designed to overcome challenges in MS lesion segmentation.
• To enhance the accuracy and reliability of MS lesion segmentation from MRI data.

Main Methods:

• A multi-layer preprocessing stage involving a Denoising GAN (DGAN-Net) for noise reduction, GAN-based methods for missing data imputation, and a Multi-level Identity GAN (MGAN) for dataset expansion.
• Utilizing a Residual Attention U-Net (RAU-Net) with identity mapping for precise MS lesion segmentation.
• Implementing techniques like identity blocks, 8-connected pixel constraints, and softened discriminator outputs to improve GAN performance.

Main Results:

• The DGAN-Net achieved peak signal-to-noise ratio (PSNR) values up to 42.21 dB for noise reduction.
• RAUM-GANs demonstrated superior performance compared to four state-of-the-art methods on the MICCAI MSSEG-2 dataset.
• Achieved a Dice score of 96.6%, Fréchet Inception Distance (FID) of 43.13, and Inception Score (IS) of 14.03.

Conclusions:

• RAUM-GANs effectively generates high-quality synthetic MRI data, enhancing robustness against noise and incomplete information.
• The framework delivers superior MS lesion segmentation performance, offering a comprehensive and scalable solution.
• Potential applicability extends to other medical imaging domains facing data quality and scarcity issues.