When a Flat Tracing Learned to Read a Stack
In the museum workshop, the bird was easy at first. One sturdy stencil, one clean outline. Then the new job landed on the bench: the same bird split across a stack of clear sheets, each holding a faint sliver. The old stencil still mattered, but it was far too flat for this.
Body scans have the same problem. A normal picture is flat, but a full scan is a pile of slices, and marking every border by hand takes expert time. Here the stencil is a picture reader trained on ordinary photos, and the clear sheets are the layered scan. If that reader already spots edges well, some of that skill can be carried over.
The first fix stayed close to home. Keep the trusted flat guide for flat jobs, and for layered scans read one slice at a time. The maker also takes a full-size rubbing before shrinking the view, so tiny edges are not lost. That extra early step helped on heart ultrasound and belly scans, doing better than other flat-only setups and getting close to the level where specialists match each other.
The second fix was clever in a different way. Turn the whole stack of clear sheets into one guide page, let the practised stencil setup read that page, then spread the guidance back through the stack. That worked well on brain tumour scans, because the reused flat core was already good at picking out patterns inside each slice.
The third fix was bolder. Build a tracer made for the whole stack, then seed it with the flat stencil copied through the depth, like making a short pile of matching stencil plates from one original. On the same brain tumour task, this fuller version did a bit better in the trickier inner parts and finished scans faster too.
That changed the old rule. A flat guide did not have to stay in flat work. It could sit inside a layered job, or be stretched into a layered starting point. That matters when fully marked scan stacks are hard to get, but strong training on flat pictures is much easier. So better outlines can come from less expert marking.