The part you can see
When you look at a 3D anatomical model, rough edges are easy to spot. A smoother version often looks more finished. That is useful for presentation, but appearance alone does not tell us whether the shape is closer to its source.
The part that needs checking
A surface edit can help in one region and change detail somewhere else. We also need to look beyond one organ: an edit can alter the space between neighbouring structures. A single average score may hide the location of a problem.
What we are investigating
Our reconstruction work compares edited models against a baseline and records where results improve or worsen. We are looking at local detail, the space between structures and the shape between sampled sections. These are separate questions, and a method needs to be assessed against each of them.
Where the work stands
The current evidence is mixed. Some comparisons show improvements, while others still show unwanted changes. Our next steps are to understand those failures and test fixed methods on new cases. The work has not established clinical accuracy or a breakthrough.
What this means for the product
We want review to be part of the workflow: the original representation, the edited model and the processing history should remain available together. A useful tool should help people find a problem as well as show them an improvement.