Can you use trackpy in cases when local position is not as important as shape of objects?
Opened this issue · 1 comments
I am looking to use trackpy for tracking lesions in a longitudinal dataset. In this dataset the lesions could move a lot between time points, and what's more important is matching of the shape of the objects between time points than necessarily displacement. Registration between time points is not possible for reasons I won't get into. Could trackpy work in such situations, perhaps where I set the search_range to a fairly large value? Any input on this?
That sounds like a neat application! If you can quantify the shapes somehow (perhaps with skimage's regionprops?), then during linking you can treat the parameters of each lesion's shape as you would treat extra spatial coordinates. Issue #470 has a basic discussion of this.
If the default Euclidean distance metric is not adequate for your shape "coordinates," you can even customize the way trackpy computes distances.