cni » projects » image-analysis » brain morphometry
differential morphometry
The term differential morphometry refers to an image analysis approach to extract differential measures from the segmented images of serial MRI. Each MRI exam is individually segmented into the major tissue classes as well as pathological anomalies. The series is then registered spatially and regional changes are evaluated automatically on a pixel-by-pixel basis. This yields exact counts not only of the total change that occurred, but also where and of what kind. For example, a trend derived from two separate measures of T2 lesion burden will be confounded by new and resolving lesions compensating each other. From a global lesion burden it looks like nothing changed, but in terms of disease activity there is in fact significant turnover (Figure 1).
Attached media: Figure 1
Examples of ‘Differential Morphometry’ applied to serial MRI of T2 lesion burden. Shown are lesion burden changes over a 1-year follow-up period for 16 different MS patients. Note that more than half of the total lesion burden captured at an individual time point expresses transient “turnover” volume of either new/enlarging or resolving lesions. Global rates of lesion burden in longitudinal studies may obscure this fact and drastically underestimate the amount of disease activity.
A schematic example of how this process is shown in Figure 2.
Attached media: Figure 2
Example schematic of the “Differential Morphometry” concept applied in development of DAI/DSI. This particular approach applies voxel-based morphometry to a series of MRI images (a) and tissue type label series segmented from each individual exam (b). White matter signal abnormalities (WMSA) are reclassified into “new”,” stable” and “resolving” lesions (c), and integrated over the entire parenchyma to yield new surrogates of disease activity. In parallel, direct changes in MRI intensity are evaluated from the intensity difference image (d). In a first approach, an empirical threshold is applied to the difference image to define an active change in tissue quality towards or away from hypo- hyper- or isointense.
a) PDw MRI series, baseline (t1) and two follow-up exams each 1 month apart (t2,t3)
b) tissue class segmentation of each exam, resulting in labels for WM (white), GM (gray), CSF(blue), and WMSA/lesions (orange).
c) differential morphometry, comparing two subsequent exam timepoints, i.e. t2-t1 and t3-t2, re-classifying all WMSA into “new/enlarging” (red), “stable” (green), and “resolving” (blue) lesions.
d) Intensity difference images of two subsequent exam timepoints, analyzing direct signal changes to support the differential morphometry results obtained in c).