Related Experiment Video
Updated: Jun 9, 2025
09:53
Role of Diffusion MRI Tractography in Endoscopic Endonasal Skull Base Surgery
Published on: July 5, 2021
3.6K
Deep learning-based super-resolution and denoising algorithm improves reliability of dynamic contrast-enhanced MRI in
Junhyeok Lee1, Woojin Jung2, Seungwook Yang2
1Interdisciplinary Program in Cancer Biology, Seoul National University College of Medicine, Seoul, 03080, Republic of Korea.
Scientific Reports
|October 25, 2024
Summary
Deep learning super-resolution and denoising (DLSD) significantly improves dynamic contrast-enhanced MRI (DCE-MRI) for diffuse glioma patients. This enhanced imaging offers better signal quality and diagnostic accuracy for blood-brain barrier leakage.
Area of Science:
- Neuroimaging
- Artificial Intelligence in Medicine
- Radiology
Background:
- Dynamic contrast-enhanced MRI (DCE-MRI) is crucial for assessing blood-brain barrier (BBB) leakage in brain tumors.
- Clinical application of DCE-MRI is limited by image noise and partial volume artifacts.
- Diffuse gliomas represent a significant challenge in neuro-oncology, requiring accurate imaging biomarkers.
Purpose of the Study:
- To evaluate the efficacy of deep learning-based super-resolution and denoising (DLSD) in enhancing DCE-MRI quality for diffuse glioma patients.
- To compare the diagnostic performance and reliability of DLSD-enhanced DCE-MRI (DL-DCE) against standard DCE-MRI (std-DCE).
- To assess the impact of DLSD on quantitative pharmacokinetic parameters and arterial input function (AIF) estimation.
Main Methods:
- Retrospective analysis of DCE-MRI data from 306 adult patients with diffuse glioma.
- Application of deep learning-based super-resolution and denoising (DLSD) techniques to std-DCE images.
- Quantitative comparison of signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and pharmacokinetic parameters between DL-DCE and std-DCE.
- Assessment of diagnostic performance using area under the receiver operating characteristic curve (AUROC) for WHO grade differentiation.
- Evaluation of AIF reliability using intraclass correlation coefficients (ICC) for parameters like Time to Peak.
Main Results:
- DL-DCE demonstrated significantly higher SNR (52.09 vs 27.21) and CNR (9.40 vs 4.71) compared to std-DCE (P < 0.001).
- DL-DCE improved the differentiation of glioma WHO grades based on the pharmacokinetic parameter Ktrans (AUC, 0.88 vs 0.83; P = 0.02).
- The parameter Ktrans derived from DL-DCE showed superior agreement and reliability compared to other AIF parameters (Time to Peak ICC, 0.79 vs 0.43; P < 0.001).
Conclusions:
- DLSD techniques substantially enhance image quality (SNR, CNR) and reliability of DCE-MRI in diffuse glioma.
- DLSD maintains or improves diagnostic performance for differentiating glioma grades using quantitative DCE-MRI parameters.
- DLSD-enhanced DCE-MRI represents a promising advancement for non-invasive assessment of BBB integrity in neuro-oncology.
