Kato and Ohno — university micro-CT tooth surfaces

Akiko Kato and Norikazu Ohno (2009). Construction of three-dimensional tooth
model by micro-computed tomography and application for data sharing.
Clinical Oral Investigations 13, 43–46.
https://doi.org/10.1007/s00784-008-0198-4
https://a-kato.agu.ac.jp/English/bottom.html
https://a-kato.agu.ac.jp/3D_tooth_models.html

Historical notice: since 2026-10-15 the app packages no geometry from these
archives. Their converted, repaired and compiled files remain in the local
model archive, and this notice keeps their attribution and the record of what
was done to them. Former placements:
46RLM1_OBJ.lzh: FDI46, right mandibular first molar → #30, mirror #19; replaced
on 2026-10-15 by ACTA-DIRECT v2 tooth 0039, a two-rooted lower first molar (the
FDI46 specimen has a third, distolingual root).
16RUM1_OBJ.lzh: FDI16, right maxillary first molar → Universal #3, mirror #14;
replaced on 2026-10-12 by ACTA-DIRECT v2 tooth 0158.
44LRP1_OBJ.lzh: FDI44, right mandibular first premolar → #28, mirror #21;
replaced on 2026-10-12 by ACTA-DIRECT v2 tooth 0217.

Each archive supplies separately registered enamel, dentin and pulp-cavity
surfaces. The project describes Shimadzu SMX-225CT acquisition; specimen-specific
scan spacing and physical coordinate calibration have not been independently
confirmed. The original coordinate scale is retained. Buccal/lingual orientation
is not independently labelled; the cervical plane is a display estimate.

Reuse status: public university research download, with no confirmed dataset
redistribution grant. These files were local prototype sources only and are no
longer packaged; permission would be needed before any app redistribution. No
Creative Commons license is asserted.

The source conversion preserves every original detail triangle, welds only
coincident coordinates for computed vertex normals, and applies one shared
proper rigid rotation/translation to the three tissues. The opposite side reflects X and
reverses triangle winding. The source conversion adds no tissue cleanup, hole
filling, synthetic canal, anatomy smoothing, nonuniform scaling or tissue
offsets.

Before the 2026-10-11 display repair, the atlas used bounded display
simplification with original source positions, locked borders and retained
source shading normals, and a separately reviewed normal-only derivative
reduced CT staircase shading in exterior detail/atlas meshes: Gaussian sigma
0.055 within radius 0.15 source coordinate units, 64 neighbors maximum,
area/normal-angle weights, change capped at 25 degrees. In that derivative all
non-normal OBJ bytes remained identical. Mirrored normals are exact reflected
original-side normals. Pulp geometry and normals are untouched. Source meshes
and original unsmoothed display files remain preserved in the local project.

CT terrace geometry, open/nonmanifold boundaries and disconnected components
are retained outside the display repair described below. In the source,
exposed occlusal dentin patches may reflect wear or segmentation; the public
low-resolution photographs do not establish which. No fine canal,
apical-foramen, expert-anatomy or clinical validation is asserted. Display colors
are illustrative, not photographic specimen color.

The separate FDI35 DXF candidate was inspected but was not selected for the
numbered atlas because of its coarse tessellation and small mesh openings.

For the former #3/#14 placement, a later shared proper Y-axis rotation of 180 degrees corrected
display orientation based on the single large inferred palatal root and paired
buccal roots. All six meshes, including pulp and atlas layers, received exactly
the same vertex/normal transform. That step changed no other records or shape. The
source FDI16/right identity remains; mesial/distal direction is not independently
source-labelled. Earlier normal-only and atlas files remain preserved upstream.

Display repair (2026-10-11): at #19/#30 until 2026-10-15 (and, until 2026-10-12,
at #3/#14 and #21/#28) the displayed meshes were derived from these sources. Small detached enamel and dentin specks were
removed, and porous or perforated occlusal enamel, with the dentin beneath it,
was rebuilt as a smooth inferred surface that follows the specimen's own porous
enamel, with at least 0.5 mm of enamel over smoothed dentin
(Tools/Models/repair_scan_display.swift). The app labels the repair; it is not
segmented anatomy. Atlas tiers were re-simplified to 12,000 triangles per tissue
from the repaired meshes. Compiled display archives also carry baked optical
attributes and smoothed display normals. The source meshes are unchanged.