I’m working with monoplanar videoradiography, and I’m trying to reduce depth ambiguity during bone tracking.
I was wondering whether SlicerAutoscoperM allows applying geometric or anatomical boundary constraints to DRR volumes.
For example, is it possible to define a condition where the femur must remain between two other volumes (or within a predefined 3D region) during registration, so that the solution space is restricted and out‑of‑plane drift is reduced?
Any insights, existing functionality, or potential workarounds would be greatly appreciated.
I’m not sure AutoscoperM exposes a hard anatomical constraint directly, but a practical workaround is to separate the problem into initialization and registration. First place the femur and the bounding volumes with landmarks or a manual transform, then run registration only in a small translation/rotation window around that pose. A cropped DRR/volume ROI can also keep unrelated anatomy from driving the metric.
For a true “must stay between these surfaces” rule, I’d log the candidate transform after each optimization step and reject or penalize poses whose transformed femur leaves the allowed region. Even a coarse parameter sweep with that test can show whether the ambiguity is a local-minimum problem or an underconstrained view. Recording the accepted pose and the boundary margins would make the result reproducible.