Research Martin Styner

Original brain images displayed on this page are provided by following source:  Original brain images for ventricles     Mono/Dizygotic twin study, Daniel Weinberger, NIMH Neuroscience      at St. Elizabeth, Washington  Ventricle segmentation      MHNCRC Imaging Lab at University of North Carolina A list of publications is located on the main page. More research can be found on the top level Research page A transparent head with lateral ventricles (smoothed data) is visualized to the right (click to enlarge image).  Animations: Ventricle + Hippocampus, Ventricle + sliced brain

Summary: Enlarged ventricular size and/or asymmetry have been found markers for psychiatric illness, including schizophrenia. However, this morphometric feature is non-specific, and its variability in healthy controls is not sufficiently understood. We studied ventricular size and shape in 3D MRI (N=20) of monozygotic (N=5) and dizygotic (N=5, same sex) twin pairs. Left and right lateral, third and fourth ventricles were segmented from high-resolution  T1w SPGR MRI using supervised classification and 3D connectivity. Segmented ventricles were rigidly co-registered. Metrics for pairwise shape similarity included binary overlap ((S intersect R)/R), Hausdorff distance, mean absolute distance and mean squared distance between object surfaces, focussing on the mean squared distance. Each object was compared pairwise with the remaining 19 objects to test if MZ or DZ twins are more alike than unrelated pairings. Volumetry could identify none of the pairs, but shape analysis resulted in a correct pairing of 6 (4 MZ, 2 DZ) twin pairs which also appeared visually as similar. The 4 remaining cases (1 MZ, 3 DZ) showed significant size difference (N=2) or asymmetry (N=2) between pairs. Additionally, we tested the hypothesis that monozygotic twins have more similar ventricular shape than dizygotic twins. Consulting the volumetric measurements, this hypothesis is rejected. But testing the shape difference measurement, we clearly detect a difference. Correcting the scale of the object with the individual volume, the analysis results in a p-value of p=0.012 for the hypothesis. Thus, using our shape measurements, we can conclude that monocygotiv twins indeed have a higher degree of similarity in ventricular shape.

The dataset is part of a Twin study dataset which has been studied before. Results of that study were published by Alycia J. Bartley, Douglas W. Jones and Daniel R. Weinberger in Brain (1997), 120, pp. 257-269 (Arcobat PDF Document)
 

Shape population - Top view of position normalized lateral ventricles
The correspondence is visualized by correspondingly colored ridges
Twin 1-5 Mono-zygotic - Twin 6-10 Di-zygotic

		*
Twin 1A-L / Twin 1B-L	Twin 1B-R / Twin 1A-R	*	Twin 2A-L / Twin 2B-L	Twin 2B-R / Twin 2A-R
**************************************	**************************************	*	**************************************	**************************************
		*
Twin 3A-L / Twin 3B-L	Twin 3B-R / Twin 3A-R	*	Twin 4A-L / Twin 4B-L	Twin 4B-R / Twin 4A-R
**************************************	**************************************	*	**************************************	**************************************
		*
Twin 5A-L / Twin 5B-L	Twin 5B-R / Twin 5A-R	*	Twin 6A-L / Twin 6B-L	Twin 6B-R / Twin 6A-R
**************************************	**************************************	*	**************************************	**************************************
		*
Twin 7A-L / Twin 7B-L	Twin 7B-R / Twin 7A-R	*	Twin 8A-L / Twin 8B-L	Twin 8B-R / Twin 8A-R
**************************************	**************************************	*	**************************************	**************************************
		*
Twin 9A-L / Twin 9B-L	Twin 9B-R / Twin 9A-R	*	Twin 10A-L / Twin 10B-L	Twin 10B-R / Twin 10A-R

 
 

SCALING INFLUENCE (each image first row MZ and second row DZ):
Top row: No Scaling
Bottom Row: Scaling normalization to uniform volume

Volume and boundary shape analysis

I computed the MSD difference between twins for the volume normalized SPHARM description. As in the volume analysis, the values of the left and right ventricles are added.  A t-test revealed that the two populations differ significantly, i.d. that the hypothesis of having the same mean could not be rejected at a significant level.

The SPHARM analysis revealed that the two populations differ in their shape similarity, despite that they do not do so regarding their volume similarity. However, the location where the object populations differ cannot be obtained using the SPHARM method. The m-rep shape analysis has the potential to provide locality.

This result might be of clinical importance. For example, it might be of interest in the analysis of discordant MZ twin studies. In a twin pair discordant for a disease, only one of the twins is sick, while the other twin healthy. If the discordant twins would have less similarly shaped ventricles, this might indicate that the studied disease manifests itself also in brain shape changes. This hypothesis can't easily be tested in a non-related population as the  biological variability would hide the expected subtle differences and only reveal large group effects.

This result suggests that the right lateral ventricle of monozygotic twins are more alike because their shapes are more similar than those of dizygotic twins.
 

		Volume difference between lateral ventricles of twin pairs
		Mean Squared Distance (MSD) between lateral ventricles of twins, that were not scaling normalized.
		Mean Squared Distance (MSD) between lateral ventricles of twins, that were scaled to normalize object volume.
Left side	Right Side

Medial shape analysis

Each one of the two proposed measures detects a higher level of similarity between MZ twins than between DZ twins at a significant p-value.  Furthermore, when comparing the twin populations to the population of non-related pairs, we observe that the similarity in MZ twins significantly differs from the similarity of non-related pairs. This is not the case for the dizygotic twin pairs, thus suggesting that there isn't a detectable difference (with our method) regarding the ventricular similarity between dizygotic twins and non-related subjects.
 
 
 

M-rep shape analysis:
Clearly the right side is significant, while the left side is not.

Left side	Right side
		Thickness P-values MZ/DZ MZ/O DZ/O Left 0.316 0.356 0.959 Right 0.016 0.011 0.646
		Location P-values Left 0.157 0.068 0.75 Right 0.026 0.005 0.91
		Thickness vs. location in 2D

The statistics and p-values shown in the tables and plots above are computed in a global analysis using cumulative values. These values were computed by integrating the individual local differences over the whole shape. A local statistical analysis is also possible in order to compute the locations where the 2 populations differ the most. The local analysis is performed for each medial atom individually and thus for these statistics the atoms are assumed to be independent of each other. This viewpoint is not fully correct, but valid in a preliminary analysis. A more thorough analysis should be done by computing the statistics using a regional kernel instead of doing it atom by atom. The local analysis was applied individually for the thickness and location difference measure.

The local analysis is visualized below. The locations of significant shape difference are not the same for the thickness and location feature.  It is clearly visible that the right side ventricle's similarity between Twins strongest differ in the front and the back using the position property of the m-reps.
 
 

Locations of significant difference between monozygotic and dizygotic twins.
The differences are shown in the frame of the common m-rep model. 
The radius and color is inverse proportional to the p-value: r,col ~1/p. All p-values smaller than 0.05 have the maximal color (red) and radius. 
Top row: side view. Bottom row: top view.

Thickness	Position
not-significant 0.10  0.05 significant
		Left side
		Right side