MRI Cross-Modal Synthesis: A Comparative Study of Generative Models for T1-to-T2 Reconstruction

TL;DR

This study compares three advanced generative models—Pix2Pix GAN, CycleGAN, and VAE—for synthesizing T2 MRI images from T1 scans, evaluating their performance on the BraTS dataset to guide model selection in clinical applications.

Contribution

It provides a comprehensive comparison of state-of-the-art models for T1-to-T2 MRI synthesis, highlighting their strengths and limitations based on quantitative metrics.

Findings

CycleGAN achieved the highest PSNR and SSIM scores.

Pix2Pix GAN had the lowest MSE.

VAE offers better latent space representation and sampling capabilities.

Abstract

MRI cross-modal synthesis involves generating images from one acquisition protocol using another, offering considerable clinical value by reducing scan time while maintaining diagnostic information. This paper presents a comprehensive comparison of three state-of-the-art generative models for T1-to-T2 MRI reconstruction: Pix2Pix GAN, CycleGAN, and Variational Autoencoder (VAE). Using the BraTS 2020 dataset (11,439 training and 2,000 testing slices), we evaluate these models based on established metrics including Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), and Structural Similarity Index (SSIM). Our experiments demonstrate that all models can successfully synthesize T2 images from T1 inputs, with CycleGAN achieving the highest PSNR (32.28 dB) and SSIM (0.9008), while Pix2Pix GAN provides the lowest MSE (0.005846). The VAE, though showing lower quantitative performance (MSE: 0.006949, PSNR: 24.95 dB, SSIM: 0.6573), offers advantages in latent space representation and sampling capabilities. This comparative study provides valuable insights for researchers and clinicians selecting appropriate generative models for MRI synthesis applications based on their specific requirements and data constraints.