A comparison between volumetric data generated by voxel-based morphometry and manual parcellation of multimodal regions of the frontal lobe

Abstract

Background and purpose: Volumetric measurements on structural MR images are an established method to investigate pathology-related volume changes in cortex. Manual volumetric methods have sometimes been referred to as the reference standard for quality control of automatic volumetric methods. While some automatic methods, like VBM, may rely on a template, manual methods use sulci as indirect landmarks for the subdivision of cortex. The purpose of this study was to compare volumetric data generated by MM and VBM on 4 multimodal regions in the frontal lobe.

Materials and methods: We investigated 4 multimodal frontocortical regions by MM and VBM in patients with frontotemporal lobar degeneration and Alzheimer disease and controls.

Results: MM and VBM results were highly correlated for dorsolateral prefrontal cortex, orbitofrontal cortex, and hippocampus, but not for the dorsal and rostral anterior cingulate. VBM results were more consistent with results from previous studies on cingulate in frontotemporal lobar degeneration. Our results may potentially be explained by 2 factors. First, the volume of small cortical regions may be more affected by anatomic variability than large regions in the MM. Second, it has been shown that the location of multimodal cytoarchitectonic areas, such as the cingulate cortex, may be difficult to predict by the appearance of sulci and gyri.

Conclusions: While both VBM and the MM may do equally poorly in predicting cytoarchitecture, the MM may add additional unrelated variance caused by anatomic variability. Thus, paradoxically, the higher anatomic precision of the MM may potentially cause a weaker relation to cytoarchitecture.

Fig 1.

The manual delineation technique in 4 regions. Top right, superior frontal gyrus; left, middle frontal gyrus. Bottom right, inferior frontal gyrus; left, dorsal anterior cingulate gyrus.

Fig 2.

Binary masks drawn on the gray matter template generated by the FSL program. In the top row, dorsolateral prefrontal cortex masks are displayed (anterior and posterior). The second row displays rostral anterior cingulate to the left, and dorsal anterior cingulate to the right. The third row displays the mask for the hippocampus, and the fourth row shows the subcallosal medial prefrontal cortex.

Fig 3.

The comparison between the manual and VBM methods for the hippocampus. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. Comparison of methods in the left (A) and right (B) hippocampus.

Fig 4.

The comparison between the manual and VBM methods for the DLPC. Gray matter intensity (from VBM method) and the volume/intracranial volume (from the manual method), are normalized as z-scores for comparison. Comparison of methods in the left (A) and right (B) DLPC.

Fig 5.

The comparison between the manual and VBM methods for the orbitofrontal cortex. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. Comparison between methods in the (A) left and right (B) orbitofrontal cortex.

Fig 6.

The correlation between the VBM and manual methods for the hippocampus. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. A, Left hippocampus (Pearson r = 0.75); B, right hippocampus (Pearson r = 0.83).

Fig 7.

The correlation between the VBM and manual methods for the dorsolateral prefrontal cortex. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. A, Left dorsolateral prefrontal cortex (Pearson r = 0.60); B, right dorsolateral prefrontal cortex (Pearson r = 0.71).

Fig 8.

The correlation between the VBM and manual methods for the dorsolateral prefrontal cortex. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. A, Left orbitofrontal cortex (Pearson r = 0.62); B, right orbitofrontal cortex (Pearson r = 0.64).

Fig 9.

The comparison between the manual and the VBM method for the DACC and RACC. Gray matter intensity (from VBM) and the volume/intracranial volume (from the manual method) are normalized as z-scores for comparison. Comparison of methods in right (A) and left (B) DACC; comparison of methods in right (C) and left (D) RACC.

Fig 10.

The anatomic variability of the cingulate sulcus in 4 sample brains from the investigated cohort. Left hemisphere is to the left and the contralateral side of the same brain is shown to the right. See McCormick et al for a definition of cingulate sulcus.