Related Experiment Video
Updated: Jan 19, 2026
05:14
Standardized Data Acquisition for Neuromelanin-Sensitive Magnetic Resonance Imaging of the Substantia Nigra
Published on: September 8, 2021
4.4K
Evaluating the Impact of Intensity Normalization on MR Image Synthesis
Jacob C Reinhold1, Blake E Dewey1,2, Aaron Carass1,3
1Department of Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, USA 21218.
Proceedings of Spie--The International Society for Optical Engineering
|September 26, 2019
Summary
Image synthesis for medical imaging benefits from intensity normalization. This study shows normalization is vital for successful deep learning-based MR image synthesis, improving results across methods.
Area of Science:
- Medical image analysis
- Artificial intelligence in radiology
- Image processing
Background:
- Image synthesis transforms input image intensity features to generate output images with different tissue contrasts.
- This technique is valuable for medical image analysis tasks like imputation, registration, and segmentation.
- Input image intensity scaling (normalization) is standard practice but optimal methods remain unknown.
Purpose of the Study:
- To evaluate the impact of various intensity normalization algorithms on image synthesis.
- To determine the optimal input scaling strategy for medical image synthesis.
- To assess the importance of normalization for deep learning-based MR image synthesis.
Main Methods:
- Compared seven different intensity normalization algorithms.
- Investigated three distinct image synthesis methods.
- Conducted experiments to analyze the effect of normalization on synthesis outcomes.
Main Results:
- Intensity normalization as a preprocessing step consistently improves synthesis results across all tested synthesis algorithms.
- Evidence suggests intensity normalization is crucial for effective deep learning-based MR image synthesis.
- The study provides insights into optimal input scaling for medical image synthesis.
Conclusions:
- Intensity normalization is a critical component for enhancing medical image synthesis.
- The findings underscore the importance of selecting appropriate normalization techniques for deep learning applications in medical imaging.
- This research contributes to advancing the reliability and performance of synthetic medical image generation.
