Cropped tissues: crop faces and cut cells

A tissue imaged in part has cells that the image boundary cuts. The mesh closes them with crop faces: triangles on the boundary, labelled like contact with the medium although nothing beyond them was observed. delaunay-watershed-3d marks them, and foambryo reads the mark from the reconstruction (dcel_mesh_from_segmentation_mask) or from a .rec or .vtk file that records it (dcel_mesh_from_file); you can also pass it yourself, DcelData(points, triangles, labels, crop_faces=mask). Then every tension method keeps the crop faces, and every junction edge and vertex that touches them, out of its equations; an interface made only of crop faces is absent from the tension map; and the pressures of the cells the crop cuts (mesh.cut_cells) are returned as nan, with a CutCellPressureWarning, because part of their surface and volume was not observed. The viewers draw the crop faces as their own surface.

mesh = dcel_mesh_from_segmentation_mask(mask)              # a mask whose cells touch the image box
mesh.cut_cells                                             # the cells the crop cuts
tensions = infer_tensions(mesh)                            # no crop interface in it
pressures = infer_pressures(mesh, tensions)                # nan for every cut cell

If the tissue really ends on an image face (a coverslip), pass physical_image_faces=("axis0_low",) to dcel_mesh_from_segmentation_mask. A mesh file written before crop faces were recorded reads them as medium contact; when it looks cropped, dcel_mesh_from_file warns (UnrecordedCropFacesWarning). The Laplace pressure methods and the uncertainty functions have no formulation for crop faces yet and raise on a cropped mesh.

What each step does with a cropped mesh:

step	on a cropped mesh
reconstruction (dcel_mesh_from_segmentation_mask, foambryo mesh)	marks the crop faces; mesh.crop_faces (one boolean per triangle), mesh.cut_cells, mesh.crop_interfaces
.rec and .vtk files	record the crop faces; dcel_mesh_from_file reads them back
every tension method	keeps crop faces, and every junction edge and vertex that touches them, out of its equations; the rest of a junction line that ends on the crop still counts
PressureComputationMethod.Variational (the default)	infers the pressures; those of the cut cells are nan, with a CutCellPressureWarning naming them
Laplace, WeightedLaplace	refuse (ValueError); has_crop_face_formulation is False
infer_tensions_with_uncertainty, compute_inference_diagnostics	refuse (ValueError)
wedge refusal (refusal=)	a crop edge is always left out of its line's mean; the policy acts on the rest of the line
viewers	draw the crop faces as see-through glass with their edges; cut cells as undetermined pressures

What it buys, measured on 433 plane crops of the 47 benchmark tissues (YoungDupre, relative error of the tensions of the interfaces that touch the crop, median over the 42 tissues at equilibrium): 13 %, 33 % and 40 % error with 10 %, 25 % and 40 % of the tissue cropped away when crop faces were read as medium contact, 10 %, 9 % and 12 % with crop faces kept out. The cut cells' pressures, reported before crop faces were recorded, had 15 % to 46 % error. On a mesh without crop faces every number is what it was before crop faces existed, bit for bit.

Example: cropped tissue.