Improved synthetic T1-weighted images for cerebral tissue segmentation in neurological diseases
| Autor: | Alexandra van Wijnen, Michelle Maiworm, Franca Petrov, Felix Rosenow, Nina Merkel, Ralf Deichmann, René-Maxime Gracien, Helmuth Steinmetz, Susanne Knake, Marlies Wagner, Esther Paule |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Adult
Male Databases Factual Computer science Biomedical Engineering Biophysics Young Adult Epilepsy Multiple Sclerosis Relapsing-Remitting Image Processing Computer-Assisted T1 weighted medicine Humans Radiology Nuclear Medicine and imaging Segmentation In patient business.industry Multiple sclerosis Brain Pattern recognition Middle Aged Cortical dysplasia medicine.disease Magnetic Resonance Imaging Female Standard algorithms Artificial intelligence Cerebral tissue Nervous System Diseases Protons Artifacts business Algorithms |
| Zdroj: | Magnetic Resonance Imaging. 61:158-166 |
| ISSN: | 0730-725X |
| DOI: | 10.1016/j.mri.2019.05.013 |
| Popis: | Structural cerebral MRI analysis in patients with neurological diseases usually requires T1-weighted datasets for tissue segmentation. For this purpose, synthetic T1-weighted images which are constructed from quantitative maps of the underlying tissue parameters such as the T1 relaxation time and the proton density (PD) may provide advantages over conventional datasets. However, in some cases synthetic images may suffer from specific artifacts, hampering accurate tissue segmentation. The goal was to improve a previously described method for the calculation of synthetic magnetization-prepared rapid gradient-echo (MP-RAGE) datasets from quantitative T1 and PD maps. Improvements comprise a B0-correction for the water-selective excitation pulses employed in T1-mapping and the use of T1-based pseudo-PD maps. Synthetic T1-weighted MP-RAGE datasets were calculated, using the standard and the improved algorithm, for 10 patients with focal epilepsy (caused by focal cortical dysplasia in 9), 10 patients with multiple sclerosis and 10 healthy control subjects and segmented with the Freesurfer toolbox. Visual inspection disclosed that segmentation of the standard synthetic datasets was inaccurate in 6 out of 10 patients with epilepsy, 7 out of 10 patients with multiple sclerosis and 7 out of 10 healthy control subjects, while the improved synthetic datasets resulted in adequate segmentation outcomes in the majority of cases. Only for one patient with multiple sclerosis and one with epilepsy, segmentation in basal temporal regions was not sufficient. Furthermore, data based on the standard algorithm showed strong signal non-uniformities in basal regions. This effect was not present in the improved synthetic datasets. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect