PhD Scientific Days 2020

Clinical Medicine III. Lectures

Multidimensional ROI-level Combination of Micro- and Macrostructural Measures for Epileptic Lesion Detection

Előadó neve

Gyebnár, Gyula, MSc

Előadó munkahelye

Medical Imaging Centre, Department of Neuroradiology

Előadó telefonszáma

06202088580

Előadó e-mail címe

gyebnargyula@gmail.com

Az előadás címe

Multidimensional ROI-level Combination of Micro- and Macrostructural Measures for Epileptic Lesion Detection

Szerző(k) neve és munkahelye

Gyula Gyebnár1
Lajos R Kozák1
1 Medical Imaging Centre, Department of Neuroradiology

Szekció

Clinical Medicine III. Lectures

Language of the presentation

English

Section, first choice

Neurosciences

Section, second choice

Clinical Medicine

Összefoglaló szövege

Introduction
In a recent publication (Gyebnar et al. 2019) we showed that the voxel-wise Mahalanobis-distance is a suitable metric of dissimilarity for the detection of malformations of cortical development (MCD), using the three dimensional distribution of diffusion tensor (DTI) eigenvalues. While the method proved sensitive to the abnormal tissue microstructure related to MCDs, specificity was constrained by the inaccuracies of spatial coregistration when comparing to control subjects. In the current work, the method was amended in two ways: a) by including cortical volumetry and morphology measures, and b) by performing the statistical analysis on the ROI-level.
Aims
The aim of the current study was to test the multidimensional approach, using the extended parameter space and ROI-level-statistics on healthy controls and individuals with MCDs; and to find the optimal parameter set to maximize the efficiency of lesion detection.
Method
Diffusion and T1-weighted imaging data of 45 healthy subjects and 13 individuals with MCDs (16 lesions) were processed with ExploreDTI and the Freesurfer software suite. Using the 360 labels of the multi-modal cortical atlas of the Human Connectome Project (HCP-MMP, Glasser et al. (2016)), ROI-level average DTI-eigenvalues, volumetry, and morphology measures were exported. The multidimensional Mahalanobis-distance was used to identify regions of abnormal tissue micro- and macrostructure using in-house software with all possible combinations of measures.
Results
MCD-related lesions were identified in 14 out of 16 cases, detection performance was improved using measures of cortical morphology (average cortical thickness, rectified mean curvature, and folding index) with reduced number of false positives. DTI-derived measures resulted in more false positives in regions usually affected by susceptibility and EPI-related distortions.
Conclusion
The surface-based approach was efficient in registering the cortical labels to each individuals’ image space, and the combination of DTI, volumetry, and morphology measures improved the detection of MCDs in the multidimensional framework.

Additional Information

Supervisor: Lajos R Kozák
lrkozak@gmail.com

Bemutatás módja

Szóbeli

Kind

Szabad

Status

elfogadva

Accepted presentation method

szóbeli

Előadás fájl jóváhagyás

nem rendelkezett róla

Előadó

155

Start

11:55

End

12:10

Authors (legacy)

Gyula Gyebnár1
Lajos R Kozák1
1 Medical Imaging Centre, Department of Neuroradiology