Evaluation of multimodal segmentation based on 3D T1-, T2- and FLAIR-weighted images - the difficulty of choosing

Tobias Lindig; Raviteja Kotikalapudi; Daniel Schweikardt; Pascal Martin; Friedemann Bender; Uwe Klose; Ulrike Ernemann; Niels K Focke; Benjamin Bender

doi:10.1016/j.neuroimage.2017.02.016

Evaluation of multimodal segmentation based on 3D T1-, T2- and FLAIR-weighted images - the difficulty of choosing

Neuroimage. 2018 Apr 15:170:210-221. doi: 10.1016/j.neuroimage.2017.02.016. Epub 2017 Feb 7.

Authors

Tobias Lindig¹, Raviteja Kotikalapudi², Daniel Schweikardt³, Pascal Martin⁴, Friedemann Bender⁵, Uwe Klose⁶, Ulrike Ernemann⁷, Niels K Focke⁸, Benjamin Bender⁹

Affiliations

¹ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
² Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany; Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
³ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁴ Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁵ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany; Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁶ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁷ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁸ Neurology and Epileptology, Hertie Institute for Clinical Brain Research, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].
⁹ Dept. of Diagnostic and Interventional Neuroradiology, University Hospital Tübingen, Hoppe-Seyler-Str. 3, 72076 Tübingen, Germany. Electronic address: [email protected].

PMID: 28188918
DOI: 10.1016/j.neuroimage.2017.02.016

Abstract

Voxel-based morphometry is still mainly based on T1-weighted MRI scans. Misclassification of vessels and dura mater as gray matter has been previously reported. Goal of the present work was to evaluate the effect of multimodal segmentation methods available in SPM12, and their influence on identification of age related atrophy and lesion detection in epilepsy patients. 3D T1-, T2- and FLAIR-images of 77 healthy adults (mean age 35.8 years, 19-66 years, 45 females), 7 patients with malformation of cortical development (MCD) (mean age 28.1 years,19-40 years, 3 females), and 5 patients with left hippocampal sclerosis (LHS) (mean age 49.0 years, 25-67 years, 3 females) from a 3T scanner were evaluated. Segmentation based on T1-only, T1+T2, T1+FLAIR, T2+FLAIR, and T1+T2+FLAIR were compared in the healthy subjects. Clinical VBM results based on the different segmentation approaches for MCD and for LHS were compared. T1-only segmentation overestimated total intracranial volume by about 80ml compared to the other segmentation methods. This was due to misclassification of dura mater and vessels as GM and CSF. Significant differences were found for several anatomical regions: the occipital lobe, the basal ganglia/thalamus, the pre- and postcentral gyrus, the cerebellum, and the brainstem. None of the segmentation methods yielded completely satisfying results for the basal ganglia/thalamus and the brainstem. The best correlation with age could be found for the multimodal T1+T2+FLAIR segmentation. Highest T-scores for identification of LHS were found for T1+T2 segmentation, while highest T-scores for MCD were dependent on lesion and anatomical location. Multimodal segmentation is superior to T1-only segmentation and reduces the misclassification of dura mater and vessels as GM and CSF. Depending on the anatomical region and the pathology of interest (atrophy, lesion detection, etc.), different combinations of T1, T2 and FLAIR yield optimal results.

Keywords: Atrophy; Automatic lesion detection; Epilepsy; Statistical parametric mapping.

Publication types

Evaluation Study
Research Support, Non-U.S. Gov't

MeSH terms

Adult
Aged
Atrophy / pathology
Brain / anatomy & histology*
Brain / diagnostic imaging*
Brain / pathology
Cross-Sectional Studies
Epilepsies, Partial / diagnostic imaging*
Epilepsies, Partial / pathology
Epilepsy, Temporal Lobe / diagnostic imaging
Epilepsy, Temporal Lobe / pathology
Female
Hippocampus / diagnostic imaging
Hippocampus / pathology
Humans
Image Processing, Computer-Assisted / methods*
Image Processing, Computer-Assisted / standards
Magnetic Resonance Imaging / methods*
Magnetic Resonance Imaging / standards
Male
Malformations of Cortical Development / diagnostic imaging
Malformations of Cortical Development / pathology
Middle Aged
Neuroimaging / methods*
Neuroimaging / standards
Young Adult