Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 30, 2026Open Access

Diffusion-Based Synthetic CT Generation from MRI: A Comparative Evaluation with Deep Learning Baselines

View Full Paper
Ask AI
Bookmark
Share

Authors

HSHuanan Su

Discussion

Loading...

Member takes

Overview

Randomized trial compares synthetic CT generation from MRI using diffusion and deep learning models, indicating superior performance.

Key Points

  • This research aims to evaluate synthetic CT image generation from MRI using various deep learning techniques.
  • Comparative analysis of deep learning models: Diffusion, U-Net, GAN, and TransformerUNet.
  • Utilized brain MRI-CT data from the SynthRAD2023 Grand Challenge dataset for experiments.
  • Results were evaluated using metrics: MAE, PSNR, SSIM, and NCC.
  • The Diffusion model outperformed other techniques in MAE, PSNR, and NCC metrics.
  • Synthetic CT images produced better anatomical structure preservation compared to real CT images.
  • The Diffusion model exhibited slower inference speed and higher GPU memory requirements.

Cite This Study

Huanan Su (2026) studied this question.

synapsesocial.com/papers/6a6af50460e2b924d3ea09f5https://doi.org/10.1051/itmconf/20268801042/pdf
View Full Paper
Ask AI
Bookmark
Share