Accuracy and Caveats

Accuracy and Caveats

Read this before you trust a number.

What is exact

• Structure-to-plate assignments come verbatim from the authors' published Index of structures — not from OCR of the plates.
• Plate AP coordinates were read from all 62 plates and match the printed series exactly: bregma +7.80 mm at plate 1, −0.35 mm per plate.
• The landmark AP offsets — lambda +4.45, interaural +7.25, occipital crest +9.95 — are the atlas's own printed figures.
• The three sources of a level are registered to each other by construction, not by fitting. Verified by re-deriving the labelled plates the same way and landing on the ones already in the file, pixel for pixel, on 61 of 62 (the remaining one is out by a single pixel).
• Moving the origin in the working frame is an exact subtraction.

What is a good approximation

• ML and DV are read off each plate's own printed coordinate box (ML ±8 mm, DV +1 to −10 mm). Midline structures land within 0.2 mm of ML 0; bilateral pairs come out symmetric.
• The regional outlines used by the SVG export land within 1.8 px — about 0.03 mm — of each plate's own drawn lines.

What to be careful with

A structure's coordinate is a label position

It is the median position of where the abbreviation is printed, close to but not the same as the structure's centre. For thin or layered structures the atlas often prints the abbreviation beside the region and points at it with a leader line — so the coordinate can sit just outside the structure it names.

The card's Label extent row tells you how tightly the figure is pinned down.

Coverage is 93%, not 100%

Labels were read from the plates automatically. 3,270 of 3,506 index entries carry at least one located label (93.3%), from 6,220 individual printed labels. 20 of the 723 structures have no located label at all, so they have no coordinate — the app says so rather than showing nothing. You may notice a region you expected is not marked.

A plate range is not a guarantee

A structure listed for a plate range is present at those levels but is not necessarily printed on every plate of that range.

The system tags are ours

The auditory / thalamus / fiber_tract chips are a convenience layer added here. They are not part of the published atlas.

Nothing here is a segmentation

The 3D views interpolate between sections that are 350 µm apart, against a plate pixel of about 17.5 µm — sampling along the brain is 20× coarser than across it. The streaking in the volume view is arithmetic, not anatomy. Most structures carry only a handful of labels (median six), so one structure reads as a sparse arc, not a shape.

The myelin plate is an adjacent section

Not the same slice as the Nissl — the two stains cannot both be applied to one section. It is aligned as the authors published it.

The skull is a different animal (experimental)

The atlas prints no skull, so the fit is computed here rather than published. It is a CT scan of a different animal, aligned on the paper's landmarks and then scaled to the drawn sections — stretched about 3% along AP and 4% across, shortened 4% in depth. Good to a few tenths of a millimetre. Context, not a surface to measure against.

For the Landmarks overlay: the AP of each landmark is exact; only the heights, and the ear canals the interaural line runs between, come off the skull fit.

The track planner is experimental

Not yet checked against a track anybody has driven. Its brain surface is the outline of the atlas's own drawn section — a fixed, sectioned brain that knows nothing about vessels, the sinus or the ventricles. An entry AP resolves only to the nearest plate.

Frame rotation is experimental

Pitch, roll and yaw are new and not fully tested. Check any coordinate they adjust against anatomy you already know. While a rotation is on, the grid, the measure tool, the projections and the 3D view stay in atlas coordinates and say so. And a plate stops being a single AP: at 17° of pitch the AP of a point drifts about 2.5 mm across the depth of the brain.

---

Cite the atlas

Radtke-Schuller S, Schuller G, Angenstein F, Grosser OS, Goldschmidt J, Budinger E (2016). Brain atlas of the Mongolian gerbil (Meriones unguiculatus) in CT/MRI-aided stereotaxic coordinates. Brain Struct Funct 221(Suppl 1):1–272. doi:10.1007/s00429-016-1259-0 — open access.

Full derivation, verification and error analysis: METHODS.md.