Article
The influence of a denoising algorithm on diffusion tensor metrics in diffusion weighted imaging (DWI)
Einfluss von Denoising diffusiongewichteter Bilddaten auf Diffusions-Tensor Parameter
Search Medline for
Authors
-
Jia Yang
- Universitätsklinikum Marburg, Klinik für Neurochirurgie, Marburg, Deutschland
-
Miriam Bopp
- Universitätsklinikum Marburg, Klinik für Neurochirurgie, Marburg, Deutschland
-
Barbara Carl
- Universitätsklinikum Marburg, Klinik für Neurochirurgie, Marburg, Deutschland
-
Christopher Nimsky
- Universitätsklinikum Marburg, Klinik für Neurochirurgie, Marburg, Deutschland
| Published: | May 8, 2019 |
|---|
Outline
Text
Objective: Diffusion weighted imaging (DWI) based fiber tractography (FT) plays an important role in pre-surgical planning. However, DWI suffers from artifacts during acquisition and low spatial signal-to-noise-ratio (SNR), which might affect diffusion metrics. Up to now, different denoising algorithms have been proposed, but there is less investigation on the influence on diffusion metrics which is the prerequisites for accurate diagnosis and FT. Position orientation adaptive smoothing (POAS) algorithm aims to improve image quality while preserving the anatomic structure and avoiding blurring. In this study, the effect of POAS on diffusion metrics is evaluated. Therefore, diffusion metrics are compared between processing using POAS and applying averaging, a common known approach to improve image quality.
Methods: 22 healthy volunteers were included in the study. For each volunteer five repetitive DWI data sets were acquired at a 3T MRI. After preprocessing applying motion and eddy current correction on the one hand, all five data sets were averaged, on the other hand all data sets were processed using the POAS algorithm. The original data sets (ORG), the POAS-processed data sets as well as the averaged (AVG) data sets were compared regarding SNR, fractional anisotropy (FA) and mean diffusivity (MD) applying tract based spatial statistics (TBSS).
Results: Averaged data revealed significantly increased SNR compared to original data, POAS-processed data resulted in significantly increased SNR compared to original and averaged data (p<0.001). TBSS analysis showed significant lower FA in POAS-processed data compared to original and averaged data across the whole brain. No significant difference in MD was seen between original and averaged data, whereas MD was altering across the whole brain. Variability of FA and DWI signal intensity in white matter was lower in POAS-processed data compared to original data.
Conclusion: Post-processing using POAS could enhance DWI image quality, decrease the variability and increase the stability of DWI signal intensity and FA value. Therefore, POAS might be capable of correcting the overestimation of FA value which was normally caused by low SNR in DWI, and providing a better estimation of FA.
