Automatisierte quantitative MRT-Analyse zur Bestimmung der epileptogenen Läsion bei Patienten mit fokalen Epilepsien
Loading...
Files
Date
Authors
Publisher
Philipps-Universität Marburg
Supervisors
Abstract
This study was designed to test a new method for automatic detection of epileptogenic
lesions in patients with focal epilepsy and known FCD. For this all patients underwent high
resolution 3T NMR of the brain. The NMR data was processed with the purpose-built
software FreeSurfer. The processing pipeline included motion correction and normalization,
the reconstruction by a VBM algorithm and the analysis and labeling of cortical and
subcortical regions.
A control group of 50 healthy adults was made up especially for this study. All members of
the control group underwent the same imaging process on the identical 3T NMR scanner
and the data was processed as described above. To guarantee a good comparability the
control group was well-balanced regarding age, gender and education. Only right handed
people were included in the study as patients as well as in the control group.
The objective was to automatically detect focal cortical dysplasias who represent
epileptogenic lesions by comparing the cortical thickness of one patient to a gender matches
control group claiming that the thickness of the patient’s cortex is significantly increased.
In addition to that blurring of the grey white boundary as indication for an epileptogenic
lesion should automatically be detected.
Statistical analysis was calculated with qdec and showed that only 60% of the patients had
significantly increased cortical thickness within the assumed area of the epileptogenic lesion.
In 50% of the patients significantly increased cortical thickness was detected that could not
be related to the patients clinical symptoms (false positive). Blurring of the grey white
boundary in the area of the assumed epileptogenic lesion could only be found in 60% of the
patients. Only in the subgroup of patients with frontal lobe epilepsy the FCD was detected
in 100% of the patients.
A subgroup analysis regarding location or side of the lesion or having underwent an
intracranial operation before did not show significant results. Anyway the group of patients
was small and there for these subgroups have almost no statistical power.
The main reason for the small number of detected lesions and the big number of false
positive detections seems to be the reconstruction algorithm. All patients included in this
59
study have a FDC. In Type I FCDs only the cortex is affected by a disordered layering so
even in high resolution NMR imaging often no correlate can be found. In addition type II
FCDs include altered cell architecture such as dysmorphic neurons or balloon-cells
(Blümcke, et al., 2012) which leads to a blurred grey white boundary. So they leave a trace
within the white matter which can be detected by T1 NMR imaging (Colombo, Salamon,
Raybaud, Ozkara, & Barkovich, 2009). For this study a volume based VBM algorithm was
used that originally was designed to identify subcortical patterns. Cortical variations and
those who are close to the cortex are frequently only detected if they leave a hint within the
white matter. So in order to detect cortical lesions as tried in this study it might be better to
use a surface based VBM algorithm for reconstruction (Thesen, et al., 2011).
In addition the small group of patients decreases the power of the conclusions that can be
made from this study. But negative impact on the results due to an unsuitable control group
can be excluded.
Review
Metadata
Contributors
Supervisor:
Dates
Created: 2014Issued: 2014-10-29Updated: 2014-10-29
Faculty
Medizin
Publisher
Philipps-Universität Marburg
Language
ger
Data types
DoctoralThesis
Keywords
MRTEpilepsieepilepsyFCDFCDFreesurferkryptogenMRIfreesurferkryptogenic
DFG-subjects
KernspintomographieEpilepsie
DDC-Numbers
610
show more
Wehrmann, Enno (1061168506): Automatisierte quantitative MRT-Analyse zur Bestimmung der epileptogenen Läsion bei Patienten mit fokalen Epilepsien. : Philipps-Universität Marburg 2014-10-29. DOI: https://doi.org/10.17192/z2014.0687.
License
This item has been published with the following license: In Copyright