Application III : Follow-up of lesions in time series of images

Figure 6: 3D visualisation of evolving lesions. The color is related to the time of appearance of each lesion.

6.9cm

The analysis of evolving processes over time, for instance multiple sclerosis lesions, tumors, or anatomical structures (i.e. cerebral ventricles) helps in diagnosis and allows the follow-up of a patient over time, especially to study the effects of a treatment. A first stage consists in aligning images of the series together with a rigid registration algorithm (). Then w We developed two different methodologies to detect and quantify evolving multiple sclerosis lesions in series of 3D MR scans (,).

• the first method constists in analyzing an apparent displacement field between two images. This vector field is computed with a non-rigid registration algorithm (). Then scalar operators (i.e. norm, Jacobian, etc.) allows to detect evolving areas ().
• the second method aims at a retrospective analysis thanks to an average model of the intensity of evolving lesion voxels. Then with a statistical inference it is possible to detect voxels with a pathological behavior.

yav++ is very useful to look at 4D results, especially because it is possible to superimpose resulting segmentations of our methods on $3D+t$ images, as it allows to change all the parameters of the $3D+t$ images interactively and makes it possible to mix two $3D+t$ images with an opacity variable.
Moreover it is possible to visually combine these results in 3D (Figure 6) to help clinicians to localize evolving lesions with respect to typical brain structures.

Figure 7: Orthogonal slicers that allow to superimpose two $3D+t$ images (6 views); (a) Slicers panel; (b1-b2-b3) image at time 0 with different opacities (c1-c2-c3) image at time 5 with same opacities

(a)

(b1)	(b2)	(b3)
(c1)	(c2)	(c3)