ORAL ATLAS / MODEL DETAIL

Denser teeth. Clearer exploration.

The latest update adds ten structures from the same Z-Anatomy source after a registration check: frontalis, orbicularis oculi (orbital and palpebral parts), platysma and the vagus nerve. Earlier updates refined 100 nerve, vessel, and lymphatic structures, added tissue materials to 104 structures, upgraded 18 teeth from source surfaces (now 20, with both lower first molars) and applied original render settings to 17 structures. Explore the source provenance, verification, and remaining resolution limits below.

The offline ZIP includes the complete model inside the HTML viewer. Choose Display tools → Download GLB to save the GLB without an internet connection.

265Individually named structures · 10 categories
4.83×Triangle count of the 20 replaced teeth
1,712,016Current total triangles · 916,323 vertices

D1–D4: learning around the atlas

The new learning workspace adds study tools without changing the anatomical meshes. It opens in English by default and retains Korean translation.

ModuleIncludedInterpretation
D1 foundations28 permanent teeth with Universal/FDI numbering, comparison at the same scale, morphology/root/surface cards, and an interactive tooth-layer schematic.Third molars remain absent. Typical morphology notes describe general anatomy and variation, not measured features of every displayed mesh.
D2 applied anatomyFive nerve pathways (the maxillary steps draw the superior alveolar, palatine and nasopalatine nerves as schematic paths), 35 muscle cards (mastication, facial expression, floor of mouth and tongue, pharynx and soft palate) with painted attachment footprints and dental connections, an innervation map for 11 nerves or branches that paints the muscles each one moves and states what an injury loses, an anesthesia map for seven standard blocks and five other techniques (local infiltration of one chosen upper tooth, Gow-Gates, Vazirani-Akinosi, AMSA and P-ASA) on either side that tints the teeth whose pulps are numbed and paints the gingiva reached on the lip/cheek or tongue/palate side, a root & canal view that cuts any tooth along its long axis and, for lower premolars and molars, draws the shortest distance in this model to the inferior alveolar nerve trunk, compares the posterior teeth of that side and gives CBCT values from patients, five stepwise clinical-anatomy lessons (local anesthesia, facial nerve branches, odontogenic infection spread, restoration, endodontics), generated neighbour, nerve-territory and innervation practice, and 3D jaw motion that opens, protrudes and moves the atlas mandible sideways, with the muscles at work and a 2D concept diagram.Only existing atlas structures are highlighted or painted. Attachments to skin, mucosa or raphes and muscles absent from the atlas mesh (tensor tympani) are named in the notes rather than drawn; styloglossus and palatoglossus appear as dashed schematic paths in the innervation map. Footprints are surface contacts in this model, not dissection measurements. Anesthesia territories are typical; variation (midline crossover, an absent middle superior alveolar nerve) is stated with each block, and the painted gingiva is a band within 2.5 mm of each tooth in this model, not a measured nerve territory. Jaw motion moves the atlas mandible with illustrative kinematics: a 35° condylar path, pure rotation for the first 20 mm, up to 15 mm of condylar slide at 45 mm, incisal and canine guidance, and the discs riding on the condyles; teeth are not collision-tested and muscles do not deform, so they are hidden in that view. Root & canal distances are measured to the nerve trunk, the longest connected piece of the atlas inferior alveolar nerve (its dental branches to each apex are left out); they describe this model, not a patient, and the third molar is not in the atlas. The maxillary sinus is a schematic outline, so upper teeth get a section but no distance. AMSA and P-ASA are shown with their described territories; an anatomical review questions how a palatal injection reaches the dental nerves.
D3 clinical reasoningThree fictional cases, each with three stages: pulpal/periapical findings, periodontal assessment and a persistent mucosal lesion. Four interactive imaging concepts (periapical, bitewing, panoramic, CBCT), plus a six-site #30 / FDI 46 chart with PD, signed GM, calculated CAL and BOP.Imaging views are authored colored schematics, not scans or images registered to the atlas. The CBCT slider changes a conceptual slice. Chart values are fictional; CAL uses the displayed signed-margin convention. A single tooth does not establish periodontitis stage or grade.
D4 comprehensive careThree fictional, three-stage cases: prognosis and phased care, anticoagulant coordination, and preference-sensitive tooth replacement. Two ordering exercises and three patient-explanation scripts with teach-back prompts and .txt export.Educational choices and model answers are conditional on each fictional scenario. They are not individualized medication instructions, validated clinical simulation, or a substitute for patient-specific assessment and supervision. Explanation exports are generic study text.
PracticeTen concept questions per year (40 across D1–D4); tooth-identification questions for all 28 atlas teeth; in D2, 3D identification of the left-side muscles, nerves and glands, each shown unnamed among three nearby structures of the same kind (10–60 mm away, up to 80 mm for an outlying target), preferring choices that do not cover the target on screen; innervation questions generated from the innervation map (what an injury to a branch weakens, which nerve or branch moves a muscle); 51 anesthesia questions generated from the anesthesia map: from its standard blocks, a site (one tooth’s pulp, or its gingiva on the lip/cheek or tongue/palate side) numbed by exactly one block of its jaw asks for that block, and an extraction asks for the unique smallest set of one or two blocks that numbs the pulp and both gingival sides; distractors never numb the site, and the answer’s territory is painted in 3D; five scenario questions ask for the other techniques (for example, a patient who cannot open the mouth needs the closed-mouth Vazirani-Akinosi block), and an extraction’s feedback names an other technique that alone numbs all three sites. In D1 and D2 the label quiz numbers the structures of eight textbook-style plates (upper and lower teeth, muscles of mastication, floor of the mouth, facial muscles, salivary glands, mandibular and facial nerve branches) and asks which structure a number marks, missed ones first, with the other numbered structures as choices. Sessions contain up to ten questions, with feedback and local missed-question review. On a phone, a question is answered from a short bar at the bottom of the viewer, so the model and its numbered labels stay visible above it; Details and sources opens the full panel.Original practice questions, not official exam items, a complete course or a competency assessment. Anesthesia answers use the map’s typical territories; site and extraction questions use only the standard blocks, so an infiltration or other technique is never marked wrong there. Browser storage is local and may be unavailable or cleared.

Study views use a controlled layout. Shift + scroll separation is suppressed only while learning is open; closing learning returns control to the normal atlas interface. English/Korean translation covers cards, lessons, cases, questions, scripts and diagram controls. D3/D4 lab values and exercise selections survive topic/language changes in page memory, but restart after refresh; no patient chart is stored.

The new illustrations are conceptual teaching graphics, not CT, micro-CT, MRI or oral scans. They are not registered to an individual tooth, included in the downloadable GLB, or calibrated for injection placement, cavity preparation, working length or joint measurements. Complete internal tooth layers and the previously listed missing structures remain absent from the atlas geometry. Illustrated disease does not modify the static atlas meshes, and no clinical imaging dataset was added.

The six staged cases, bilingual summaries, illustrations, scripts and questions were written for this explorer using the linked references as factual support. No source illustrations or examination question banks were imported. Source links inside each lesson identify further reading, not endorsement or clinical validation.

Study references: ADA tooth numbering · Confederation College anterior tooth morphology · posterior tooth morphology · Washington State University oral anatomy · Texas Tech HSC El Paso anatomy · AAE root canal anatomy · ADA restorative treatment guidance. D3/D4 references include ADA imaging guidance, AAP staging and grading, and AAE endodontic diagnosis. Additional topic references appear in the viewer.

Four distinct kinds of improvement

MethodWhat changedWhat it means
Replacement with source tooth surfacesDenser published BodyParts3D 3.0_95 surfaces were registered to the existing teeth.This uses additional surface information from the published source, beyond subdividing the existing triangles.
Original render subdivisionThe render levels already authored in the Z-Anatomy Blender file were applied to 17 structures.This smooths the surface. It does not add newly measured anatomical tissue or microstructures.
Source-preserving curve and capsule refinementDenser native nerve/vessel curves and two Catmull–Clark subdivision levels for lymphatic surfaces refine the existing mesh.Existing source anatomy remains the basis. More triangles do not establish new branches, internal tissues, or scan accuracy.
Color, lighting, and procedural surface detailDisplay settings and illustrative tissue textures help distinguish structures and surface curvature.This does not represent a photographic reproduction of measured tissue color, moisture, microtexture, or optical properties.

The initial GLB contained 668,156 triangles; the current model has 1,723,532, an increase of approximately 157.95%. Triangle count describes surface density, not biological accuracy. Sources and comparison values for each structure are recorded in structures.json.

Teeth: additional source-surface information for 20 structures

The replaced teeth increased from 32,192 to 155,352 triangles in total. Each tooth was registered using rotation, translation, and uniform scaling, without nonuniform deformation or ad hoc sculpting. Left-side teeth follow the existing atlas’s bilateral symmetry. Crown/root material regions were transferred from the existing surface; they do not represent separately measured internal tissue layers.

Counts below are per tooth. Each row represents two structures, one on each side.
ToothPrevious trianglesCurrent trianglesSource ID
Lower first molar · both sides (outer crown and root surface)1,8368,818FMA55705
Lower first premolar · both sides1,2225,900FMA55694
Lower second molar · both sides1,6768,102FMA55706
Lower second premolar · both sides1,1805,624FMA55695
Upper canine · both sides1,4706,966FMA55798
Upper first molar · both sides2,17610,786FMA55698
Upper first premolar · both sides1,6848,054FMA55689
Upper lateral incisor · both sides1,4546,842FMA55680
Upper second molar · both sides1,8449,254FMA55697
Upper second premolar · both sides1,5547,330FMA55688

For the lower first molars, only the original outer crown and root surface (1,836 triangles) was registered and replaced (RMS 0.017 mm, 95th percentile 0.035 mm, sampled maximum 0.097 mm, bounding difference 0.014 mm); their original dentine surface and inner enamel shell stay unchanged inside it, so both molars keep the dentine distinction. The other eight original teeth were checked again: the upper and lower central incisors match except at incisal edges that Z-Anatomy reshaped (up to 0.44 and 0.59 mm), and the lower lateral incisors and canines were rescaled and reshaped (0.57 and 0.66 mm), so they keep their original surfaces. BodyParts3D 3.0 has no tongue or third molars; the only reachable BodyParts3D 4.0 tongue is coarser (502 triangles) and differs by RMS 2.1 mm, so the tongue is unchanged. The other 8 teeth retain their existing surfaces. This is not a model that fully separates every tooth into enamel, dentine, pulp, and periodontal ligament layers.

Registration checks

We compared 12,000 area-weighted surface samples in each direction. The following are the largest values measured for any direction across the 20 accepted teeth.

MetricUpper value, rounded upward
Root-mean-square distance (RMS)0.019924 mm
95th-percentile distance0.041844 mm
Maximum sampled distance0.128253 mm

These values describe agreement with the existing atlas surface. Error against an actual patient and clinical accuracy were not measured. A sample-based check does not establish the maximum error across the entire continuous surface.

Sources: Kevin Mattheus Moerman’s BodyParts3D STL conversion · Original BodyParts3D 3.0_95 data · Original BodyParts3D publication

Original render settings for 17 structures

These structures are the gingiva, soft palate, oropharynx, both parotid and accessory parotid glands, both sensory and motor roots of the trigeminal nerves, both sphenomandibular and stylohyoid ligaments, and both temporomandibular joint capsules. The source’s render-subdivision settings smooth their surfaces. No new body scan was added.

The 8 teeth that were not replaced and the tongue still retain their original surfaces. The surface refinement below retains the prior geometry of the four trigeminal nerve roots.

Geometry refinement and lossless delivery

100 structures received geometry refinement: 80 native curves (26 nerves, 36 arteries, and 18 veins), plus 20 lymphatic structures including tonsils. The curves retain their source control points, handles, radii, branches, and world transforms. Two Catmull–Clark subdivision levels smooth the lymphatic surfaces. These changes refine existing anatomy; no new branches or measured microanatomy were added.

132 structure geometries remain unchanged from the previous version, including all 28 teeth and the four trigeminal nerve roots. The roots receive UV texture coordinates and tissue materials without geometry changes. In total, 104 structures receive the tissue-material and UV update.

The downloadable GLB is 21,876,824 bytes and uses EXT_meshopt_compression. A compatible viewer or importer is required. Compression uses no quantization or lossy filter; floating-point attribute values and triangle winding were verified unchanged after decoding. On the website, oral-atlas.html (about 2 MB) fetches the same gzip data as oral-anatomy.glb.gz, showing progress and letting the browser cache it; the model URL carries a hash of its content (?v=…). Once the model has loaded, the website registers a small service worker (sw.js) that keeps the page and the model: later visits open at once and work without a connection, pages are still fetched fresh when online, and the GLB and ZIP downloads are not cached. The viewer in the offline ZIP embeds that data and includes its decoder, so base viewing remains offline without a service worker. Compression reduces storage size; it does not increase anatomical accuracy.

Named nerve branches, sphenoid bone and nearby structures

Seven named branches per side are now separate, selectable structures. Z-Anatomy draws them as unnamed splines inside larger nerve curves, so each name is an interpretation of the spline’s course:

  • Infraorbital nerve: the maxillary nerve from the first point within 1 mm of the maxilla, where it reaches the orbital surface (approximating the inferior orbital fissure), through the infraorbital groove and canal, with its four terminal splines at the infraorbital foramen.
  • Incisive nerve: the spline running forward inside the mandible from the mental-foramen junction to the midline, with the four dental twigs leaving it toward the canine and incisors.
  • Facial nerve branches: temporal, zygomatic, buccal, marginal mandibular and cervical groups, assigned by where each spline ends (above the zygomatic arch, lateral orbit, lips and nose, lower border of the mandible, neck). The trunk is divided where the cervical branch leaves it, and its distal part forms the marginal mandibular branch. The intracranial and intratemporal course and the posterior twigs stay with the facial nerve.

Every original triangle remains exactly once, in the parent nerve or one of its branches; only flat caps close the trunk division points. All other structures, including their vertex and index data, are unchanged.

Sphenoid bone. The original Z-Anatomy mesh (3,650 vertices, 7,308 triangles) is added with bone and suture roles, so the pterygoid muscles now have their attachment bone. The denser BodyParts3D 3.0 surface (FMA52736, 15,322 triangles) was registered with one rotation, uniform scale and translation, but it differed from the atlas surface by RMS 0.73 mm, 95th percentile 1.37 mm and maximum 7.6 mm. That exceeds the 0.2 mm gate used for the teeth, so it was not used.

Nearby structures. For each structure the explorer lists up to 16 neighbours within 8 mm, computed once from this model. About eight area-weighted samples per mm² (at most 60,000 per structure) give the closest sampled distance; the contact area is the mean, over both directions, of the sampled surface within 0.5 mm of the other surface. At least 20 mm² is shown as broad contact and at least 1 mm² as partial contact. In single view, the explorer finds, for the inspected structure only, the vertices within 0.5 mm of each listed contact (fading to none at 0.75 mm) and paints them in that neighbour’s legend color; this is computed in the browser from the same surfaces. These describe the atlas geometry, not clinical measurements.

Ten structures added from the source after a registration check

Frontalis, the orbital and palpebral parts of orbicularis oculi, platysma and the vagus nerve (X), on both sides, now come from the same Z-Anatomy Blender file as the rest of the atlas. A structure was added only if it passed every step of this check:

  • Same frame: eight reference structures read from that file (the mandible, both maxillae, a zygomatic bone, both buccinators, depressor anguli oris and the sphenoid) coincide with the GLB within 0.006 mm after the fixed axis conversion, so the new meshes need no fitting.
  • Closed surfaces: each muscle is a closed (watertight) surface.
  • No buried tissue: at most 0.1% of each new muscle surface lies more than 0.5 mm inside a bone, no more than the existing facial muscles (up to 3.0%).
  • Contacts: each muscle meets the structures it attaches to or overlies in this model (frontalis meets orbicularis oculi; the orbital part of orbicularis oculi meets the zygomatic bone and maxilla around the orbit; platysma meets the mandible and the depressors of the lip and mouth corner).

Frontalis and platysma extend beyond the region of this atlas; they are clipped at its box and capped, so the model bounds and centre, saved links and the schematic layer stay unchanged. The vagus nerve is built like the other atlas nerves (bevel depth × point radius, 20 radial and 18 per-segment samples); compared with the existing glossopharyngeal and hypoglossal nerves built the same way, radii agree within 0.4% and every source spline is covered. Its trunk is cut where it leaves the atlas region. The ten structures add 51,917 vertices and 91,094 triangles; every earlier primitive is byte-identical. Styloglossus and palatoglossus have no mesh in any source and are drawn as schematic paths; the levator and tensor veli palatini were later added from BodyParts3D 4.0 (below). Nearby lists were recomputed with tools/nearby.

Levator and tensor veli palatini from BodyParts3D 4.0

The two muscles of the soft palate that Z-Anatomy and BodyParts3D 3.0 lack were added on both sides from BodyParts3D 4.0 (Database Center for Life Science, CC BY 4.0), using the copy published by the human-atlas project, which simplified each structure with meshoptimizer within 0.2% relative error. A structure was added only if it passed every step of this check:

  • Registration: BodyParts3D 4.0 is not in the atlas coordinate frame, so one similarity transform (scale 0.9831, 0.34° of rotation beyond the fixed axis conversion) was fitted jointly on nine neighbouring structures: the sphenoid, both maxillae, both temporal bones, both palatopharyngeus muscles and both superior pharyngeal constrictors. Each then lies within a median 0.24–0.49 mm of the atlas surface (RMS 0.37–1.23 mm). The palatine bones were left out of the fit because their BodyParts3D 4.0 shapes differ (RMS 2.2 mm).
  • Closed surfaces: each muscle is a closed (watertight) surface.
  • No buried tissue: 0–2.2% of each new muscle lies more than 0.5 mm inside a bone, within the range of the existing muscles around it (0–4.2%).
  • Contacts: tensor veli palatini meets the sphenoid and the soft palate; levator veli palatini lies within 0.4 mm of the temporal bone and ends about 1.4 mm short of the atlas soft palate, so its muscle card lists the soft palate as not touching in this model.

The model bounds and centre, saved links and the schematic layer are unchanged, and every earlier primitive is byte-identical. The four structures add 5,766 vertices and 11,516 triangles. Nearby lists were recomputed with tools/nearby. Both muscles have cards in the muscle study, and the innervation map paints levator with the pharyngeal plexus and tensor with the V3 motor branches.

Schematic layer for structures the source does not model

Display tools → Schematic paths (off by default) adds 22 schematic items, eleven per side. The source atlas has no meshes for them, so tools/schematic/derive.mjs draws each one from structures it does contain, and the result is stored in src/schematic-data.json. These items are not part of the GLB, are not counted among the 265 structures and are not used for nearby lists or contact areas.

  • Posterior superior alveolar nerve: from the maxillary nerve in the pterygopalatine fossa, along the posterior superior alveolar artery over the maxillary tuberosity, to the second and first molar root apices.
  • Middle superior alveolar nerve (inconstant): from the infraorbital nerve in its groove, through the lateral sinus wall 1 mm inside the bone surface, to the premolar apices.
  • Anterior superior alveolar nerve: from the infraorbital nerve 7 mm before the infraorbital foramen, which is located where the nerve's terminal branch splines leave it, through the anterior wall to the canine and incisor apices.
  • Greater and lesser palatine nerves: from a point 3.5 mm below the maxillary nerve, standing in for the ganglion, along the descending palatine artery and then the greater palatine artery to the canine region; the lesser nerve turns back into the soft palate.
  • Nasopalatine nerve: from the same point across to the septum, which is not modelled, then through the incisive canal to the palate behind the central incisors.
  • Lingual and sublingual arteries: from the external carotid artery opposite the tip of the hyoid greater horn, above the hyoid, lateral to genioglossus and inside the tongue to its tip. The sublingual branch runs to the sublingual gland.
  • Styloglossus: from the tip of the styloid process, taken as the upper end of the stylohyoid ligament, forward, down and medially to the side of the tongue above hyoglossus, then forward along the tongue's lateral border (1 mm inside its surface).
  • Palatoglossus: from the oral surface of the soft palate down the palatoglossal arch, 5 mm in front of the palatine tonsil, into the side of the tongue. Both muscles have no mesh in Z-Anatomy or BodyParts3D; the innervation map draws styloglossus with the hypoglossal nerve and palatoglossus with the pharyngeal plexus.
  • Maxillary sinus: the hollow of the maxillary body inside a 1.5 mm wall. The zygomatic and palatine bones close the openings of the maxilla surface. The sinus reaches down to the premolar–molar apices, and the infraorbital canal ridges its roof. It measures 4.0 mL per side, smaller than a typical adult sinus, because this atlas's maxillary body is small; it is not enlarged beyond the bone.

Paths are dashed. Where bone or other tissue hides them, they show faintly through it. Hovering names an item and marks it as schematic. Measuring has three modes: two-point distance, the angle at the second of three points, and the shortest distance between two structures—found by comparing sampled vertices of both surfaces, then measuring every vertex within 4 mm of that pair exactly against the other surface's triangles, both ways—with the stored contact area when they touch. Note pins store their structure, a point in the assembled model frame (mm) and up to 140 characters in the link (note=…, at most 12). Each release runs axe-core's WCAG 2.2 A/AA rules in several interface states and a keyboard pass in the browser check. The local anesthesia lesson shows the superior alveolar and palatine paths for the upper teeth. Run node tools/schematic/derive.mjs after changing the model; npm run check fails if the stored data no longer matches the GLB.

Maxillary artery: two named alveolar branches

Besides its trunk, the source maxillary artery contains eight unnamed spline tubes per side, all over the posterolateral maxilla above the molars and premolars. They were grouped by how their ends connect:

  • Middle superior alveolar artery (3 splines): one spline leaves the infraorbital artery in the infraorbital groove (1.2 mm from it) and descends in the lateral wall of the maxillary sinus to end beside the first molar and second premolar. Two short twigs leave it. This branch is inconstant in people.
  • Dental branches of the posterior superior alveolar artery (5 splines): one spline leaves the end of the maxillary artery trunk at the pterygomaxillary fissure and descends over the maxillary tuberosity to the first and second molar region. Four twigs leave it, two of them close to the end of the existing posterior superior alveolar artery.

Whole splines were moved; none was cut, every original triangle is kept exactly once and all other primitives are byte-identical. The maxillary artery keeps its 51 mm trunk. The source has no names for these splines, so the names interpret their course. The nearby lists were recomputed with tools/nearby.

Lateral temporomandibular ligaments: side names corrected

The atlas frame has +x toward the patient’s left, and every paired structure lies on the side its name gives, except the two lateral temporomandibular ligaments as received from the source. The mesh named left (.l) lay on the patient’s right, its bounds centre at x ≈ −53.2 mm, and the mesh named right (.r) on the left at +53.2 mm. Their neighbours confirm the swap: the mesh named left touched the mandible and the right temporal bone, joint capsule, deep part of the masseter and parotid gland, and lay 0.7 mm from the right articular disc; none of its 16 listed neighbours was on the left. The mesh named right did the same on the left. The articular discs of these joints are built the same way, one mesh placed twice, and there the left disc is the mirrored copy; for the ligaments the mirrored copy was the one named right.

Each ligament now carries the name of the side it lies on. The two named GLB nodes exchanged their placement (translation, mirroring and mesh reference), so names, Korean labels and descriptions stay with their side; the mesh data and every primitive are byte-identical and the GLB keeps its size. Their bounds were exchanged in structures.json and the nearby lists were recomputed with tools/nearby; beyond the side, the ligaments’ listed distances and contact areas change only by sampling (at most 0.1 mm and 3 mm²). npm run check now reads the GLB and fails if any .l structure lies at x ≤ 0 or any .r at x ≥ 0, or if the stored bounds or the English and Korean side labels disagree with the model.

Surface detail: illustrative tissue appearance

The default Realistic tissue display gives every structure an illustrative tissue color and procedurally generated surface texture, and soft contact shading (screen-space ambient occlusion) shows where surfaces meet. Atlas colors restores the familiar category colors: yellow nerves, red arteries, blue veins and green lymph nodes. Select a structure and use Surface close-up for a closer inspection. These are display choices, not newly measured microscopic anatomy. The downloadable GLB embeds 12 PNG texture maps: color, normal, and roughness maps for each of the four tissue groups (nerves, arteries, veins, and lymphatic tissues). Shading for bone, teeth, muscle, mucosa, glands and joints is computed in the viewer from 3D noise anchored to each mesh; it needs no UV map and is not stored in the GLB. Normal maps and bump shading change lighting at the surface; they do not add anatomical structures.

  • Bone: an ivory cortical surface with broad mottling and scattered small pores. Foramina, sutures and canals come only from the original geometry.
  • Teeth: enamel shifts from a warmer cervical third to a translucent blue-grey incisal or occlusal edge, with faint horizontal growth lines (perikymata) in close-up; roots use a yellower cementum tone. Fissures, pits and root grooves are shaded from the curvature of each tooth mesh, so occlusal detail is clearer on the higher-resolution teeth than on the original lower-resolution ones. Enamel, dentine and pulp thicknesses are not modeled.
  • Muscles, tendons and ligaments: streaks suggest fascicles; tendons and ligaments are silvery and fibrous. Most muscles follow their mesh’s dominant direction. Orbicularis oris and the pharyngeal constrictors use a circular layout; temporalis, genioglossus and depressor anguli oris a fan; buccinator and mylohyoid a fixed textbook direction. These are simplified layouts, not measured fibre orientation.
  • Tongue, gingiva and mucosa: the tongue dorsum shows a median groove, a fine filiform texture and scattered fungiform papillae in close-up while the ventral surface stays smooth; attached gingiva shows stippling; other mucosa is moist with faint vascular mottling. This is surface shading only; papillae and taste buds are not added as geometry.
  • Salivary glands: irregular lobules separated by faint septa. Lobule size and arrangement are illustrative.
  • Nerves: fine directional grain suggests the outer connective-tissue sheath. The exterior remains continuous; it is not rendered as bare, braided bundles of nerve fibers.
  • Arteries and veins: restrained wall texture and highlights improve surface readability. They do not add capillaries, new branches, or separately measured wall layers.
  • Lymph nodes: a subtle fibrous capsule texture suggests their outer covering. Internal lymphoid follicles are not represented as bumps on the capsule.

The existing Z-Anatomy atlas supplies the nerve and vessel routes, branches, and anatomical placement. Denser curve tessellation and capsule smoothing refine the existing representation; they do not supply a new scan or prove additional anatomical accuracy. No hilar indentation is invented for individual nodes because its orientation has not been measured. Surface grain, tissue colors, and optical properties are illustrative, and have not been calibrated against a particular patient. Texture detail cannot recover the missing microanatomy listed below.

Anatomical references for these illustrative choices: Aparicio & Monje: human nerve tissue · University of Leeds: vessel walls · Columbia University: lymph-node capsule and internal follicles. These sources describe general anatomy; they are not source scans or a clinical validation of the rendered textures.

Browse and inspect individual structures

  1. Choose Browse structures. Anatomy shows the assembled model.
  2. Structures appear on ten category boards without overlapping. Left/right pairs sit together and their relative physical sizes are preserved.
  3. Use the category selector to focus on one group. Search in the Explore drawer and use All / Visible and Left / Right filters.
  4. Select a structure or its name to rotate it in single view. Single view and the 3D views in D1–D4 learning turn freely in any direction, like a trackball, so you can roll a tooth over its crown or root; a view button (Front, Left, Right, Top, 3D) or leaving the view sets it upright again. Return restores the previous browsing position.
  5. Expand Technical details in the inspector for dimensions, triangle counts, and source-detail information. Use Display tools for visibility, opacity, and clipping.
  6. Choose Anatomy to return to the assembled positions.

The board uses frontal orthographic projection. Each complete structure is moved as a unit; its surface is not broken into pieces. Calculations checked cell and surface-bound overlaps, adjacency of left/right pairs, label clearance, and preservation of structure size. Clipping does not create internal tissue: while a section is on, the inside of each cut structure (its back faces seen through the cut) is drawn in a flat, slightly darker base colour so the section reads as filled. This needs no closed meshes (150 of the 265 structures are closed; the mandible and maxillae are not) and adds no geometry. Current view · oblique cuts square to the viewing direction at the moment it is chosen or Align is pressed; the plane then stays fixed in the model while the view turns, the slider moves it through the model along that direction, and a copied link stores the direction (section=view,50,keep,x,y,z). Selected tooth · long axis cuts the last selected tooth cheek to tongue along its long axis: the principal axis of the tooth surface (area-weighted), pointed from crown to apex, with the plane square to the mesiodistal direction given by the neighbouring teeth along the arch; the slider moves the cut from the mesial to the distal surface (section=tooth,50 with that tooth selected).

Labels (Display tools) name the structures that are most visible in the current view. Every shown mesh is drawn into a small offscreen image in a colour that encodes its structure; faded and hidden structures are left out and the section cut is respected. The largest visible regions are labelled, and each leader line ends at the pixel of its region farthest from any other structure. Platysma and frontalis are drawn at 22% of the muscle opacity in the assembled view by default, because these thin sheets otherwise cover the neck and forehead; hover and clicks reach the structures behind them, and superficial=1 in a link shows them fully. Search also accepts older Sino-Korean names, English synonyms and abbreviations, and Universal/FDI tooth numbers; the labels keep the standard Korean terms. The label quiz uses the same method on a plate: only the plate’s structures are numbered, a lower size threshold and a finer offscreen image keep thin nerve branches and ducts, the numbers read down the left column and then the right, and they stay fixed once placed while the view turns.

Separation works in two stages. Up to 60%, every structure moves straight out from the model centre, without turning, to 3.2 times its centre's distance from it at 60%, and the camera keeps its viewing angle; a little separation therefore opens the gaps while each structure keeps its place relative to its neighbours. From 60% to 99%, each structure moves on from its spread position to its slot in the category board (eased, so the movement starts and ends gently) while the view turns from its angle to the front. At 100% the category board above takes over, with its labels and pages. Copy presentation link (Display tools) saves the current view with present=1: the link opens with the 3D model filling the window and the panels hidden, for slides or a projector; Esc or Exit presentation returns to the normal layout.

Separate university references, opened online

The base oral model works offline. University reference models is an optional feature that opens other specimens in their providers’ public viewers. These references are neither included in the base GLB nor combined as if they came from the same person.

Reference specimenProduction and resolutionProvider terms
Pharynx and Floor of MouthUniversity of Dundee. Created by Eve Laws, with anatomical review by Luke Reid. CT reconstruction and digital sculpting. Provider metadata lists 1,037,531 faces.CC BY-NC-SA 4.0 · external embed.
Upper Left Second MolarQueen Mary University of London · MuCAT / Graham Davis. Micro-CT of an actual extracted tooth. Provider metadata lists 789,151 faces.Shown only through a public embed. No download license was verified, so no reuse permission is granted here.
Lymphatic system: anatomy of a lymph nodeE-learning UMCG. Educational cross-section model made for medical-student teaching; no oral-node scan provenance is supplied. Provider metadata lists 191,842 triangles.CC BY-NC-SA 4.0 · separate external embed; no geometry included in this GLB.

Histology notes

The inspector’s Histology section describes what the selected structure’s tissue looks like under the light microscope, in English and Korean, as a few short points: for the teeth enamel, dentine, pulp, cementum and periodontal ligament; the gingiva, tongue and taste buds, soft palate, pharynx, epiglottis, the parotid, submandibular and sublingual glands and their ducts, the palatine tonsil and lymph nodes, the temporomandibular disc and capsule; the regions of the brain (six-layered neocortex with the motor, somatosensory and visual cortex, insula, entorhinal cortex, hippocampus, amygdala, striatum, globus pallidus, thalamus, hypothalamus, pineal gland, choroid plexus, ependyma, white matter, cerebellum, midbrain, pons and medulla), the pituitary, thyroid and suprarenal glands, thymus, spleen and spinal cord; the organs — heart wall and valves, trachea, bronchi and lung, pleura, nasal mucosa and paranasal sinuses, oesophagus, stomach, small and large intestine, appendix, liver, gallbladder, bile and pancreatic ducts, pancreas, peritoneal folds, kidney, ureter, bladder and urethra, testis, epididymis, ductus deferens, seminal gland, prostate, erectile tissue, ovary, uterine tube, uterus, cervix and vagina, and the cornea, sclera, iris, lens, retina, vitreous body, eye chambers, tarsal plates, lacrimal gland and passages, tympanic membrane and auditory tube; and the general tissues — bone and the jaw bones, hyaline, elastic and fibrocartilage, intervertebral discs, joint capsules, bursae and tendon sheaths, tendons and ligaments, fasciae, skeletal and facial muscle, elastic and muscular arteries, veins and dural sinuses, peripheral nerves, ganglia and the optic and olfactory nerves. Notes are chosen by structure name and category (src/histology.js): every structure of the head and mouth page and of both body pages has one (109 notes). The female page’s peritoneal “ligaments” get the peritoneal-fold note, its round and ovarian ligaments a fibromuscular note and the round ligament of the liver and ligamentum venosum a note on fetal remnants. The notes are general textbook histology written for this atlas — not specimen images and not measured from the 3D models. Twelve labelled schematic drawings, drawn for this atlas as SVG in src/histology-diagrams.js and labelled in the page language, accompany the main notes: tooth and periodontium (also for the gingiva and jaw bones), the six layers of the neocortex (cortex, motor, somatosensory and visual cortex, insula), cerebellar cortex, alveolar wall and blood–air barrier, nephron and collecting duct, classic liver lobule, small-intestinal villi and crypts, osteon, skeletal muscle fibre and sarcomere, muscular artery and vein, peripheral nerve and the layers of the retina. They are simplified and not to scale. Thirty light micrographs from Wikimedia Commons illustrate 36 notes — tooth (ground section), compact bone, visual cortex, cerebellum, lung, trachea, liver, small and large intestine, oesophagus, stomach, pancreas, gallbladder, cardiac and skeletal muscle, peripheral nerve, artery and vein, kidney tubules, hyaline and elastic cartilage, lymph node (pig), thyroid (horse), parotid gland, tongue, testis, epididymis, prostate, ovary, endometrium and pineal gland. Each keeps its own licence (CC0, public domain, CC BY or CC BY-SA); the viewer shows the author, licence and source under the image, and the licence file lists every image. They were resized to at most 720 px and re-encoded as JPEG (public/histology/), are included in the offline package, and are fetched only when a note with an image is shown.

Male and female body pages

Two website pages show the body with the same viewer: the whole male body from the top of the skull to the toes (with the brain from BodyParts3D, without the inner ear), and the whole female body from the top of the skull to the feet, with the female arms shown by their bones. Each has its own GLB and metadata (pelvis-male.glb, pelvis-female.glb with pelvis-male-structures.json, pelvis-female-structures.json); the head and mouth model is unchanged. On the website each page fetches its model in four parts cut from that GLB without changing any vertex or triangle — bones and joints first, then organs, muscles and fasciae, and vessels and nerves — and draws the parts already read while the rest downloads; the view can be turned once every part is in, and Download GLB gives the complete file. Structure names and descriptions are written in English only for now.

Whole male body — Z-Anatomy

Read directly from the Z-Anatomy Startup.blend (Blender 3.5, CC BY-SA 4.0), the same file as the head atlas (SHA-256 9f08a17e…afcd), with the same fixed axis conversion. 2,486 structures of the whole body except the brain and inner ear (the brain comes from BodyParts3D, below), inside a region box (x ±340 mm, y 0–1,720 mm, z −140 to 145 mm), none of them cut: 1,549 original meshes and 937 Bézier curves. The head — the frontal, parietal, occipital, temporal, sphenoid and ethmoid bones, the maxillae, palatine, zygomatic, nasal, lacrimal and inferior concha bones, the vomer and the mandible, the ear ossicles, the ethmoidal cells and frontal and sphenoidal sinuses, the nasal cartilages and mucosa and the nasopharynx; 28 teeth; the jaw joints and the sphenomandibular and stylomandibular ligaments; the facial and scalp muscles, the masseter, temporalis and pterygoids, the genioglossus and hyoglossus, the six extraocular muscles and levator palpebrae with the common tendinous ring and trochlea, the tarsal plates and the temporal and masseteric fasciae; the pituitary; the eyes (sclera, cornea, anterior and posterior segments and anterior chamber, iris, lens, vitreous body and retina) with the suspensory ligament and the lacrimal gland, canaliculi, sac and nasolacrimal duct; the tympanic membrane and auditory tube; the tongue, gingiva, soft palate and uvula, the parotid, accessory parotid and sublingual glands and the parotid and submandibular ducts; the olfactory, optic, oculomotor, trochlear, trigeminal (with its roots and the ophthalmic, maxillary and mandibular branches to the teeth, tongue and chin), abducens, facial (with the chorda tympani) and vestibulocochlear nerves; the cerebral arteries of the arterial circle with their cortical and striate branches, the basilar artery with the cerebellar and pontine arteries, the ophthalmic, maxillary, temporal, occipital, labial and nasal arteries; the dural venous sinuses, cavernous sinus and basilar plexus, the ophthalmic veins and the veins of the face and scalp; the facial, parotid, auricular, mastoid and occipital nodes and the palatine tonsils; the neck — the hyoid, the cervical vertebrae C1–C7 with their discs and the nuchal ligament, the larynx (thyroid, cricoid, arytenoid and corniculate cartilages, the epiglottis, the thyrohyoid, cricothyroid and quadrangular membranes and the intrinsic laryngeal muscles), the pharyngeal constrictors, stylopharyngeus and palatopharyngeus with the oropharynx and laryngopharynx, the thyroid and submandibular glands, the platysma, sternocleidomastoid, suprahyoid, infrahyoid, scalene, prevertebral and suboccipital muscles, the upper trapezius, levator scapulae and the deep back muscles of the neck, the investing layer of the cervical fascia, the common, internal and external carotid, facial, vertebral and transverse cervical arteries, the internal and external jugular, facial, retromandibular and vertebral veins, the glossopharyngeal, vagus, accessory and hypoglossal nerves and the submental, submandibular, jugulodigastric, deep and superficial cervical and retropharyngeal nodes; the thorax — ribs and costal cartilages, sternum, thoracic vertebrae, the joints and ligaments of the ribs and spine, the intercostal muscles and diaphragm, the heart with its chambers, valves and papillary muscles, the lungs in lobes, the pleura, trachea and bronchi to the segmental bronchi, the great, coronary and pulmonary vessels, the thymus, oesophagus, vagus nerves and sympathetic trunks and the mediastinal lymph nodes; the pectoral girdle and both upper limbs — clavicle, scapula, humerus, radius, ulna, the eight carpal bones, metacarpals and phalanges, the shoulder, elbow, radio-ulnar, wrist and hand joints with their capsules, ligaments, bursae and tendon sheaths, the pectoral, shoulder, upper back, arm, forearm and hand muscles, the fasciae, septa and retinacula, the axillary, brachial, radial and ulnar vessels to the palmar arches and digital vessels, the cephalic, basilic and median cubital veins, the brachial plexus with its axillary, radial, musculocutaneous, median and ulnar nerves to the fingers, and the axillary and cubital nodes; the abdomen — stomach, duodenum, liver, gallbladder and bile duct, pancreas, spleen, suprarenal glands and ureters, the omenta and colon, the abdominal wall and back, the abdominal aorta, inferior vena cava, coeliac, mesenteric, renal and portal vessels and nodes; and the pelvis and both lower limbs: the pelvic organs, bones, joints, muscles, vessels, nerves and lymph nodes, the femur, patella, tibia, fibula, the bones of the feet, the knee, ankle and foot joints with their ligaments, menisci, bursae and tendon sheaths, the thigh, leg and foot muscles, the fasciae, septa and retinacula, and the femoral, popliteal, tibial, fibular and plantar vessels and nerves. Meshes keep their original vertices and triangles; closed surfaces are oriented outward. No mesh is clipped: the box holds the whole body. Vessels, nerves, ducts, bronchi and the small intestine are tubes built like the head atlas's (bevel depth × point radius, 20 radial and 18 per-segment samples), keeping only the samples needed for the tube path and radius to stay within 0.05 mm of the sampled curve (this removes most of the tube vertices without changing the paths); none is cut. Genital, urinary, digestive, heart, lung, lymph-node, bursa, meniscus and fat-pad meshes are low-polygon in the source; the viewer draws them with PN-triangle smoothing through their original vertices (display only). 3,220,243 vertices and 5,945,227 triangles; 54.4 MB as gzip.

The head is read with the rest of the body from the same file and frame, so the vessels and nerves of the neck now continue into the skull. Its structures form three more categories — Eye & ear (with the ear ossicles and the tarsal plates), Teeth and Head & face muscles — and a Head & face preset; the nose, sinuses and nasopharynx join Lungs & airways, the tongue, gums, palate and salivary glands Digestive organs, and the pituitary Endocrine glands. Z-Anatomy’s brain and inner ear are left out for their licences. The Z-Anatomy model notice lists the “Brainder” and “White matter” brain models among its sources with no licence and does not mark which of its brain objects came from them, so no brain object is used: not the cortex, white matter, basal ganglia, diencephalon, brainstem, cerebellum, ventricles, cranial-nerve nuclei, the falx and tentorium or the pineal gland (the oral page has none either); the brain shown comes from BodyParts3D (below). The cochlea and vestibule, and the one cochlear nerve mesh beside them, come from “Anatomy of the Inner Ear” (University of Dundee, CC BY-NC-SA 4.0) and are left out like the kidney. The cerebral arteries of the arterial circle, the dural venous sinuses and the cranial nerves, which are Z-Anatomy curves, are shown; the cranial nerve and foramina work of the University of Dundee that Z-Anatomy adapted is CC BY 4.0 and credited in the licence file. The epicranial aponeurosis, the frontalis, occipitalis and temporoparietalis and the temporal and masseteric fasciae start faded with the other sheets. In the head the source has no wisdom teeth, no separate intrinsic tongue or palatal muscles and no semicircular canals; the zonular fibres have two faces or fewer and are left out, and so are two unnamed septal arterial curves, the drawing helpers of the eyeball (equator, meridians, ciliary-body and zonular curves), a schematic central-canal curve and the path curve of the medulla. The middle cerebral artery segments, spelled “M3 segment” and “M3-segment” on the two sides in the source, are named M1, M2 and M3, and the two lateral temporomandibular ligaments, which carry each other’s side suffix in the source, are named by their position. The teeth and jaws are the source surfaces; the head and mouth page has registered higher-resolution teeth and keeps the dental learning modes. The source has no cervical plexus, ansa cervicalis, recurrent or superior laryngeal nerve and no superior thyroid or lingual artery as separate objects, and its thyroid gland is a low-polygon surface of the right lobe only. The parathyroid glands are flat nine-vertex placeholders (0.15–0.25 mm thick) and are left out. The source's “ascending part” of the trapezius is its upper part from the occiput and nuchal ligament; it is named by position as the descending part (the lower part, which the source calls descending, is the ascending part). The suprarenal glands and the thyroid form an Endocrine glands category. The platysma and the investing cervical fascia start faded with the other fasciae.

Brain — BodyParts3D 4.0

The brain is not taken from Z-Anatomy (see above) but from BodyParts3D 4.0 (isa_BP3D_4.0_obj_99, DBCLS, CC BY 4.0), whose elements are each one original surface: 74 structures — the superior, middle and inferior frontal, orbital, precentral, postcentral, supramarginal and angular gyri, the front and back of the superior temporal gyrus, the middle and inferior temporal, fusiform, parahippocampal and cingulate gyri, the insula and the occipital lobe, the cerebral white matter, corpus callosum, fornix and its commissure, hippocampus and amygdala, the caudate nucleus, putamen and globus pallidus, the thalamus with its stria medullaris, the hypothalamus and the pineal gland (with the Endocrine glands), the lateral, third and fourth ventricles, cerebral aqueduct and choroid plexus, the midbrain and cerebral peduncles, the pons, medulla oblongata and cerebellum (the midline structures in the two halves BodyParts3D has). Vertices and triangles are unchanged; closed surfaces are oriented outward. BodyParts3D is a different body in its own frame, so the brain was placed by one similarity transform (uniform scale 0.974, rotation 0.06°) fitted by ICP of the BodyParts3D frontal, parietal, occipital, temporal and sphenoid bones onto the page’s Z-Anatomy bones (median surface distance 0.09–0.22 mm, 0.13 mm overall); the same transform carries every brain surface, so the brain keeps its own shape and its fit to its own skull. Checks after placement: at most 5.3% of any brain surface lies within the skull bones, no deeper than 5 mm (gyral crowns against the inner table and the pons against the clivus, where the two bodies differ); the basilar artery lies on the pons (0–0.1 mm), the pituitary meets the hypothalamus and the middle cerebral artery runs over the insula. The tentorium cerebelli was left out: its straight-sinus edge lies 11.7 mm from the page’s Z-Anatomy straight sinus. Not shown: the cerebellar lobules and vermis, the brainstem and cranial-nerve nuclei and the falx cerebri; the sulci are the spaces between the gyri, not objects. The brain is a category and preset of its own (with the arteries, veins and nerves), loads with the organs and is drawn in its atlas colour. With the brain, the HRA spinal cord and the three schematic structures below the page has 2,593 structures: 1,654 meshes and 939 tubes, 3,478,170 vertices and 6,414,881 triangles; 58.8 MB as gzip.

In the upper limbs Z-Anatomy numbers the fingers (first finger of hand to fifth finger of hand); the record names keep them and the labels say thumb, index, middle, ring and little finger. The lateral and medial cords of the brachial plexus are not separate objects in the source: they are drawn as the beginnings of the musculocutaneous, median and ulnar nerves (only the posterior cord is its own curve). Left out: the radioscaphocapitate, triquetrocapitate and ulnotriquetral ligaments of the wrist (two flat faces each, like the conoid ligament and the sternoclavicular and acromioclavicular discs) and the intertubercular tendon sheath and the tendon sheaths of the flexor carpi radialis, flexor pollicis longus, extensor pollicis longus and extensor digiti minimi (no faces). The small wrist and hand ligaments are simplified low-polygon surfaces of 8–33 vertices and the shoulder and elbow bursae are simplified 178-vertex surfaces in the source; the bursae are shown with the same curved smoothing as the other bursae. The deltoid, brachial and antebrachial fasciae and the dorsal fascia of the hand start faded with the other fasciae.

In the thorax and abdomen the source labels the parts of the trapezius the other way round (its lower part from the thoracic spines is called descending); they are named by position, so the lower part is the ascending part and the upper part the descending part. The left iliocostalis cervicis has no side suffix in the source. The source draws the oesophageal plexus and the vagal trunks, which reach the stomach left of the midline, as part of the right vagus nerve; its left vagus ends at the aortic arch. The liver is shown as the source's whole-liver surface with its peritoneal ligaments; the source's eight Couinaud segments overlap that surface by 1–3 mm and would show doubled, so they are not shown. The kidneys, the renal pelves and the intrarenal arteries and veins are left out: the Z-Anatomy licence notice lists its kidney as adapted from Kidney by Lissie Cowley under CC BY-NC 4.0, whose non-commercial term cannot be offered under this page's CC BY-SA 4.0, so no part of it is used and the renal arteries and veins end at the hilum (the head atlas leaves out the same kidney for the same reason). Two openly licensed kidneys were tested in its place. The BodyParts3D 4.0 kidneys, after a similarity fit of BodyParts3D T11–L3 and the 11th and 12th ribs onto the page (vertebrae 1.0–1.8 mm), lie 15–20 mm from the left renal vessels and run up to 12 mm into the right psoas and the duodenum; moved to the vessels they enter the suprarenal glands. The HRA Visible Human Male kidneys (v1.3), after a fit of T11–L2, overlap the suprarenal glands (10–19 % of their surface) and the duodenum. Both lie 18–22 mm in front of the quadratus lumborum, so neither is shown. Not in the source: the pericardium and the phrenic nerves (drawn as schematic structures, below), the splanchnic nerves and the thoracic duct; the spinal cord exists only as a schematic 5 mm cross-section of its tracts and is not shown. The spinal cord shown instead is the HRA Visible Human Male cord (spinal-cord-male v1.1, CC BY 4.0, 30 segments C1–S5, original triangles): each HRA vertebra (VH_M_Vertebrae v1.2, same frame) was fitted to the page vertebra of the same name by a similarity ICP with its neighbours (mean surface distance 1.2–2.3 mm), every cord vertex moved with a blend of the two vertebrae bracketing its height, and the cord was then centred in the page's dural sac slice by slice (median shift 5.0 mm, at most 8.1 mm). It reaches from C1 to the conus medullaris at L1–L2, 1.6 mm from the top of the cauda equina, and stays clear of the vertebrae (0.9 mm); the Z-Anatomy dural sac is narrower than the cord at the cervical enlargement and the upper thoracic levels, where the cord passes it by up to 4.0 mm. Left out: the conoid ligament, the intertransverse ligaments of the thoracic spine and the articular discs of the sternoclavicular and acromioclavicular joints (no usable faces), The fasciae, the chest and abdominal walls (pectoralis major, intercostal muscles and membranes, transversus thoracis), the pleura and the greater omentum start faded with the earlier layers so the organs stay visible.

Schematic structures (not from a source model). Neither Z-Anatomy nor BodyParts3D has the pericardium or the phrenic nerves, so they are drawn for this atlas — the only structures on the page that come from no source model. The pericardium is a closed sac about 3 mm outside the heart chambers and the roots of the ascending aorta, pulmonary trunk, venae cavae and pulmonary veins, bridged across the clefts between them and pressed in around the lateral pericardial nodes (1 mm voxel offset, closing, Gaussian smoothing, marching cubes, quadric simplification to 25,920 triangles); it starts faded with the pleura. Each phrenic nerve is a 1.1 mm-radius tube along a smoothing spline through waypoints: on the front of the anterior scalene (lateral border above, medial border below), behind the subclavian vein and in front of the subclavian artery, lateral to the origin of the internal thoracic artery, beside the brachiocephalic vein and superior vena cava (right) or over the left side of the aortic arch (left), then on the lateral surface of the schematic pericardium, in front of the lung root, into the diaphragm beside the caval opening (right) or the cardiac apex (left). They are marked Schematic in the selection details and hover names and flagged "schematic": true in the metadata; their shape and course are illustrative, not measured.

Names follow the source with sides made regular (Left/Right testicular artery and vein become Testicular artery.l and so on). Four bowel surfaces are named by position and shape: the source's Sigmoid colon runs from the end of the sigmoid loop at about S3 to the pelvic floor and ends at the level of the external anal sphincter, so it is shown as the rectum (its shape and size also match the BodyParts3D rectum); Descending colon includes the sigmoid loop, Ascending colon includes the caecum (one blind end in the right iliac fossa), and the curve Jejunum runs from the upper left abdomen to the ileocaecal junction and is shown as the jejunum and ileum. Z-Anatomy calls the toes "fingers of foot"; the record names keep the source names and the labels say great toe, second to fourth toe and little toe. Left out from the source: the zona orbicularis (edges without faces), the intercornual ligaments (two flat faces each) and the triradiate cartilage (a growth cartilage, fused in the adult); the anterior and posterior talofibular, calcaneofibular, dorsal calcaneocuboid and plantar calcaneonavicular (spring) ligaments, which are flat four-vertex placeholders; and the posterior tibiotalar, lateral talocalcaneal and intercuneiform ligaments and the tendon sheath of the extensor hallucis longus, which have no or too few faces. Many other small knee, ankle and foot ligaments are simplified low-polygon surfaces in the source (some are boxes of 6 to 12 vertices). The fasciae (including the crural and popliteal fasciae), the flat abdominal muscles and the small intestine start faded (8% opacity) so the structures beneath stay visible; hover and clicks reach the structures behind them.

Not in the source: the anal canal, the perineal muscles (bulbospongiosus, ischiocavernosus, transverse perineal muscles, external urethral sphincter), the perineal membrane, the scrotum with the coverings of the spermatic cord, and the patellar ligament and quadriceps tendon (the source has them only as helper objects for its labels). Two other sources were checked and not used. BodyParts3D 4.0 adds nothing here: its rectum and puborectalis are the surfaces Z-Anatomy already has (as "sigmoid colon" and pubo-analis), and its "superficial perineal muscle" is a mirrored copy of the external anal sphincter. The VOXEL-MAN segmented internal organs of the Visible Human Male (CC BY 4.0) have a scrotum and an ischiocavernosus; after a similarity fit on the hip bones, sacrum and coccyx (mean surface distance 3–4 mm) the Visible Human testes lie 41–49 mm from the Z-Anatomy testes, and after a local fit on the ischiopubic rami and crura 41% of the ischiocavernosus lies inside the Z-Anatomy bone and erectile tissue, so neither passed the registration check.

Female pelvis — HRA 3D reference organs

From the official CCF 3D Reference Object Library of the HuBMAP Human Reference Atlas (CC BY 4.0; github.com/hubmapconsortium/ccf-3d-reference-object-library, commit f1a3a63): Visible Human Female v1.2 organ files (pelvis, vertebrae, uterus, vagina, ovaries, uterine tubes, ligaments of the uterus and ovaries, urinary bladder, ureters, large intestine, small intestine) and the v1.4 blood vasculature. All share one frame (+X patient left, +Z anterior; 0.795 m below the united-body frame in y), so no registration was needed. 109 named parts keep their original vertices, triangles and normals (except a few moved out of the muscles, below); with the thorax and upper abdomen below, the colon, jejunum, aorta, inferior vena cava and mesenteric vessels that were cut at the earlier top of the region are now whole. The hip bone is modelled as compact and trabecular bone of the ilium, ischium and pubis on each side; the trabecular parts lie inside and show in a section. The small intestine and the ascending and descending colon start faded (15% opacity) so the uterus and ovaries stay visible.

Names come from the HRA node identifiers, with sides taken from position. Three corrections: the two round ligaments are labelled the other way round in the source (its "right" lies on the patient's left) and are named by position; the part the source calls the abdominal ostium of the uterine tube lies at the uterine horns and is named as the uterine openings of the tubes; the vein the source labels "superior rectal artery" (node VH_F_superior_rectal_vein, vein material) is named as the vein. The source model has six lumbar vertebrae; the lowest is shown as L6, with L4 and L5 above it. The three sigmoid arteries and veins keep the source's letters a to c. Not in the HRA source: the external genitalia, urethra, anal canal, pelvic floor, perineal and abdominal-wall muscles, pelvic nerves, lymph nodes and the iliac and other parietal arteries; the HRA v1.4 vasculature has no vessels of the lower limbs. The muscles and the lower-limb bones come from the University of Denver models below.

Female thorax and upper abdomen — HRA 3D reference organs

From the same library and frame: the v1.2 heart, lung, liver, gallbladder, biliary tree, pancreas, spleen, kidney, thymus, vertebrae, spinal cord, large and small intestine files and the v1.4 vasculature (228 parts). The heart in chambers, septum, five papillary muscles and four valves; the lungs in their twenty bronchopulmonary segments with the hila, the trachea, carina, main, lobar and segmental bronchi and the tracheal and bronchial cartilages; the thymus; the liver by its capsule with the falciform, coronary, left triangular and round ligaments, the ligamentum venosum and the hepatoduodenal ligament; the gallbladder, hepatic, cystic and bile ducts, the hepatopancreatic ampulla and its sphincter, the main and accessory pancreatic ducts (the source's ventral and dorsal ducts); the pancreas (head, neck, body, tail, uncinate process); the five parts of the duodenum, the transverse colon and both colic flexures; the spleen in its five surfaces; the kidneys with fibrous capsule, cortex, renal columns, pyramids and hilum; C7, the twelve thoracic and the upper three of the six lumbar vertebrae of the specimen; the spinal cord in segments from C8 to S4; and the ascending aorta, aortic arch and its branches (each joined from two source pieces), thoracic aorta, venae cavae, pulmonary trunk, arteries and veins, coronary arteries and cardiac veins, and the coeliac, hepatic, portal, splenic, colic and renal vessels. The source's liver segments and surface areas coincide with the liver capsule, and its hilum subdivisions duplicate the hilum surfaces, so they are not shown; the laryngeal cartilages are shown with the neck (below). No part is cut. Not in the HRA source: the stomach and oesophagus, the ribs, sternum and shoulder girdle (they come from the BoneHub CT, below), the diaphragm, pericardium and pleura, the chest, abdominal-wall, back and shoulder muscles, the nerves other than the spinal cord and the lymph nodes.

Female hip, thigh, leg and foot — University of Denver models

The muscles, the lower-limb bones, the joint cartilage and the knee ligaments come from the final 3D models of the Visible Human Female by the Center for Orthopaedic Biomechanics, University of Denver (CC BY 4.0; doi:10.56902/COB.vh.2022.1; Andreassen et al., Sci Data 10, 34, 2023), segmented from the same body as the HRA organs and already checked by their authors for overlaps. 124 parts, 62 on each side: obturator internus and piriformis (pelvic muscles); iliacus, psoas major, the gluteal muscles, gemelli, quadratus femoris, obturator externus, pectineus, the adductors, gracilis, sartorius, rectus femoris, the vasti, tensor fasciae latae, semitendinosus, semimembranosus and both heads of biceps femoris (hip and thigh muscles); tibialis anterior, extensor digitorum and hallucis longus, fibularis longus, both heads of gastrocnemius, soleus, plantaris, popliteus, tibialis posterior and flexor digitorum and hallucis longus (leg muscles); the femur, patella, tibia, fibula, the seven tarsal bones and one surface for the metatarsals and phalanges (the DU file named Phalanges also contains the metatarsals); the cartilage of the femoral head, acetabulum, distal femur, patella, tibial condyles, distal tibia and talus; and the anterior and posterior cruciate and the tibial and fibular collateral ligaments. The DU leg muscles end at the ankle: their tendons in the foot, the calcaneal tendon and the plantaris tendon are not in the source, and the descriptions say so; the source has no fibularis brevis or tertius and no muscles or ligaments of the foot. The DU feet are plantarflexed, as in the specimen. They are placed in the HRA frame by one similarity transform fitted between the DU hip bones, sacrum and coccyx and the HRA compact hip-bone layers, sacrum and coccyx: scale 1.041, mean surface distance 1.6 mm (left and right hip bones), 1.9 mm (sacrum) and 3.1 mm (coccyx); the DU hip bones, sacrum and coccyx themselves are not shown; the DU leg and foot bones are, since the HRA has none. Because the two sources segment the bones separately, thin muscles and the acetabular cartilage overlapped the HRA bones in places; those vertices were moved to 0.1 mm outside the bone with a smooth falloff (at most 1.9 mm). A few HRA parts passed through the DU muscles — the right suspensory ligament of the ovary through the psoas, the internal pudendal and inferior rectal veins through the obturator internus (on whose surface they run in the pudendal canal), short stretches of the cardinal ligament, common iliac vein and bowel — and were moved out of the muscles the same way (at most 11 mm). The hip joints then fit without overlap (femoral head and acetabular cartilage). The DU leg muscles, cartilage and knee ligaments come from the same segmentation as the DU bones shown with them and do not enter those bones (checked by ray parity; only the femoral-head cartilage reaches up to 0.6 mm into the DU femur), so they were not moved. The correction was repeated after the thoracic and upper lumbar vertebrae were added: the upper psoas was moved off the new HRA lumbar vertebrae (at most 1.8 mm), and where the right kidney passed through the psoas it was moved out of the muscle (at most 3.4 mm). The region (x ±195 mm, y −910 to 625 mm, z −190 to 60 mm) now contains every part whole.

Female knee — HRA knee parts on the DU knees

The DU models have no menisci, patellar ligament or quadriceps tendon. They come from the HRA VH_F v1.2 knee files (VH_F_Knee_L/R, VH_F_Ligaments_Knee_L/R, VH_F_Muscles_Knee_L/R): the medial and lateral menisci (one mesh per knee in the source, split into its two pieces and named by position), the patellar ligament, the tendon of the quadriceps femoris and the anterolateral ligament of each knee — 10 parts. The HRA knees are posed about 6 cm further apart than the DU legs, so each knee was fitted separately by a similarity transform between the HRA distal femur, patella, proximal tibia and fibula and the DU bones of this page (left: scale 0.981, rotation 8.0°, mean surface distance 1.0–1.3 mm; right: scale 0.984, rotation 11.0°, 1.1–1.4 mm); the HRA knee bones and their ligaments, which the DU models already have, are not shown. The fitted parts were moved out of the DU bones and knee cartilage in the same way (at most 3.4 mm); where the menisci lie between the femoral and tibial cartilage, overlaps of at most 0.3 mm (mean 0.03 mm) remain. Not in either source for the female lower limbs: the vessels and nerves, the fasciae, the joint capsules and bursae, the ankle and foot ligaments, the fibularis brevis and tertius, the muscles of the foot and the leg tendons below the ankle.

Female neck, shoulders, arms and hands — HRA and BoneHub CT bones

The neck comes from the same HRA files and frame: C1 to C6 (VH_F_Vertebrae), the spinal cord segments C1 to C7 (VH_F_Spinal_Cord) and the thyroid, cricoid, arytenoid, corniculate and epiglottic cartilages of the larynx (VH_F_Lung), 20 parts with their original vertices, triangles and normals. In this specimen the larynx and trachea lie about 1 cm right of the midline; the paired cartilages keep the source sides, which match their position in the larynx. The region now reaches the skull base (x ±250 mm, y −910 to 725 mm, z −190 to 70 mm); no part is cut. The HRA has no female arm, and no open source has the muscles, vessels or nerves of a female upper limb, so the arms are shown by their bones. They come from BoneHub, Visible Human Full-Skeleton 3D Bone Models (S. H. Alavi and M. Asseln, University of Twente; CC BY 4.0; doi:10.57967/hf/10464), segmented from the University of Denver aligned CT of the same body: 71 bones — on each side the clavicle, scapula, humerus, radius, ulna, the eight carpal bones, the five metacarpals and the phalanges of each finger (one surface per finger in the source), the twelve ribs — and the sternum (one surface for its three parts). One similarity fit of the BoneHub C7 and thoracic vertebrae onto the HRA vertebrae of the page placed them (scale 0.997, mean surface distance 0.8 mm; 1.2–2.4 mm from L1 to the sacrum, which were not fitted). The CT surfaces are finely tessellated; each bone was simplified within 0.05 mm of its surface (meshoptimizer; 1,916,120 to 421,150 triangles), the kept vertices keeping their positions and the normals of the full surface. As the BoneHub documentation notes, the CT does not fully cover the elbows, so the distal humeri and the proximal radii and ulnae are incomplete (said in their descriptions), and the costal cartilages are not segmented, so the ribs end short of the sternum. Not in any source for the female neck: the hyoid, the thyroid and parathyroid glands, the pharynx and oesophagus, the neck muscles and vessels (only the roots of the aortic-arch branches) and the cervical nerves.

Female head, breasts and intervertebral discs — BoneHub skull and HRA reference organs v2

The newer HuBMAP reference organs (cdn.humanatlas.io/digital-objects/ref-organ, CC BY 4.0) share the frame of the files above — their larynx, trachea and spinal cord coincide with those on the page — so they are placed without any fit: the mouth (v1.0; upper and lower teeth as one surface each, gingiva, hard palate, mucosa of the soft palate, three surfaces of the tongue, fungiform and vallate papillae, buccal mucosa, floor of the mouth, the labial frenula and the parotid, submandibular and sublingual glands), both eyes (v1.3; 23 parts each, from the sclera, cornea, choroid and retina with the macula, fovea, optic disc and ora serrata to the iris, pupil, ciliary body, muscle and processes, zonular fibres, lens, aqueous humour, vitreous body, conjunctiva, trabecular meshwork and scleral venous sinus), the intervertebral discs from C2–C3 to L6–S1 with their nuclei pulposi (v1.0, named by the two vertebrae each joins), both mammary glands (v1.1; lobes, lactiferous ducts and sinuses, nipple, areola, areolar glands, suspensory ligaments and fat), the greater omentum, the epiploic appendages of the transverse colon and the renal pelves (v1.0), and the palatine tonsils (v1.2). They were simplified within 0.05 mm (meshoptimizer; 4,548,099 to 1,170,098 triangles for these and the brain). The skull is the BoneHub CT surface of the cranium with the maxillae (one surface) and the mandible, placed by a similarity fit of BoneHub C1–C3 onto the page's HRA C1–C3 (mean surface distance 0.6–0.8 mm); with that fit the HRA upper teeth sit on the maxillae (median 2.4 mm) and the eyes in the orbits. The mandible is slightly open in the CT, so it was closed onto the HRA lower teeth and gingiva by one rigid fit (5.7°; median distance of the lower teeth from the mandible 11.4 → 3.0 mm), which leaves the condyles 0.2–0.4 mm from the mandibular fossae. The brain is the Allen Human Reference Atlas brain of an adult female donor (brain-female v1.4) as the HRA places it in this body. Against the CT skull 9.5% of its cortex lay beyond the inner table (up to 5.7 mm, mostly the occipital lobes and cerebellum), so it was fitted into the cranial cavity by one similarity transform (scale 0.919, rotation 0.4°, translation 2.5 mm); afterwards 0.7% of the cortex samples lie beyond the inner table, mostly where the bone is thin or open. Every Allen region was checked against where its name puts it (its lobe, or inside its parent — the thalamus for the thalamic nuclei, the amygdala for its nuclei, the putamen for the pallidum): 268 of 276 passed; the frontal pole (lying in the back half of the hemisphere), the frontomarginal gyrus, the atrium of the lateral ventricle (at the front) and the posteroventral putamen (high near the midline) failed on both sides and are left out, and six regions of six to nine vertices (olfactory bulb, anterior olfactory nucleus, central canal) are not shown. The file labels the Allen hemispheres the other way round, so the regions are named by position; midline structures split in two (corpus callosum, anterior commissure, ventricles, aqueduct, vermis, pineal body, midline thalamic nuclei) are named as left and right halves. The HRA skin (v1.5) is posed with the arms raised and longer legs than the CT and DU bones and is not shown. The lower gingiva, which in this source takes in the mucosa down to the vestibule, the buccal mucosa, the fat of the breasts and the greater omentum start faded. With these the female page had 970 structures.

Female muscles, organs and vessels from the CT — TotalSegmentator

The Visible Human Female was scanned by CT before she was frozen and sectioned, and the National Library of Medicine publishes those images (normal CT, exam 370: 985 slices 1 mm apart from the top of the head to the upper thighs, pixels 0.49–0.94 mm). We resampled them to a 1 mm grid and segmented them with TotalSegmentator 2.18 (Wasserthal et al., 2023; code and the open-licence tasks used here under Apache-2.0) on the CPU: the tasks total, headneck_muscles, head_muscles, headneck_bones_vessels, oculomotor_muscles and abdominal_muscles. The abdominal-muscle model, which covers the trunk from about T4 to L4, ran in three blocks overlapping by 100 mm and joined in the middle of each overlap, so that no block edge decides a muscle.

Because the CT is of the same woman, its bones can be fitted to the bones already on the page. The segmented bones were fitted by similarity fits per region — head (cranium and C1–C2) 1.3 mm mean surface distance, neck (C3–T1) 0.7 mm, thorax (thoracic vertebrae, ribs, sternum) 1.0 mm, lumbar spine 1.4 mm, pelvis 2.5 mm, shoulders 1.2–1.3 mm, femora 1.2–1.3 mm — and every structure was carried by a blend of these fits weighted by its distance to each region. The lumbar vertebrae were fitted as one group, because the model numbers this woman's six lumbar vertebrae wrongly. As a check, organs the page already has, segmented from the CT and carried the same way, lie 3–6 mm from the page's surfaces (spleen 3.6 mm, aorta 3.9 mm, inferior vena cava 3.7–4.7 mm, kidneys 4–6 mm, liver 3.9–6.5 mm, heart 5–6 mm): the organs moved somewhat between the scan and the freezing.

The page's structures were voxelised at 1 mm. Where a carried structure entered one of them it was trimmed back (contact within 1 mm is allowed), and a structure was kept only if at most a quarter of its volume overlapped the page and no overlap was deeper than 8 mm, and for a pair only if both sides passed. 74 of 96 passed: the oesophagus, stomach and laryngopharynx; the maxillary and frontal sinuses; the thyroid and both suprarenal glands; the hyoid and the cartilages of the sixth to tenth ribs (the upper costal cartilages were not found in this CT); the right common carotid and subclavian arteries, the internal carotid arteries and the brachiocephalic and internal jugular veins; the optic nerves; the temporalis, lateral pterygoid, four recti and inferior oblique of each side; the sternocleidomastoid, platysma, three scalenes, sternothyroid and prevertebral muscles, and the middle and inferior pharyngeal constrictors; the trapezius, levator scapulae, pectoralis major, serratus anterior and latissimus dorsi; and the rectus abdominis, external and internal oblique, quadratus lumborum, erector spinae and transversospinalis. The sinuses are air spaces inside the CT skull, whose surface on this page encloses them, so for the air spaces the skull itself was not counted as overlap, and at least half of each had to lie inside it (frontal sinuses 100 %, maxillary sinuses 86–90 %). Left out: the masseter and medial pterygoid (overlapping the HRA parotid gland and cheek lining), the superior oblique and levator palpebrae (the eye and orbit), the thyrohyoid (the larynx), the external acoustic meatus (the parotid), the common and external iliac arteries (the HRA iliac veins and the DU psoas), the nasal cavities and nasopharynx (the HRA soft palate and tongue), the pericardium (tried as the outer 2 mm of the segmented sac, which lies within the HRA heart), the digastric and oropharynx (incomplete) and the superior pharyngeal constrictor (nothing left after trimming). The segmentation of the rectus abdominis ends at L5, that of the erector spinae and transversospinalis at T3 and that of the serratus anterior at T4; their descriptions say so.

Surfaces were extracted by marching cubes from the lightly smoothed masks, smoothed (Taubin, 30 iterations) and simplified within 0.2 mm (meshoptimizer; 3,234,414 to 169,028 triangles); the abdominal-muscle masks were first smoothed along the body axis so that their cross-sections change gradually across the joins of the blocks, and holes of up to 6 cm² in the latissimus dorsi and trapezius sheets were closed. They are grouped as jaw and eye, neck, shoulder-chest-back and trunk muscles and endocrine glands; the superficial neck, chest, back and abdominal-wall muscles start faded so that the organs stay visible. The female page now has 1,044 structures: 3,026,346 vertices and 6,012,662 triangles, 54.9 MB as gzip.

For both pages the nearby-structure lists use the same sampling as the head atlas, and npm run check verifies that each page's metadata, sides, categories, faded layers and GLB contents agree. These are general reference anatomies, not patient scans.

Remaining limits

The tongue still uses the original 343-vertex, 682-triangle surface. The viewer draws it with curved PN-triangle smoothing that passes through every original vertex (at most 0.62 mm, on average 0.12 mm, from the original faces); this is display only and the GLB is unchanged. No invented microscopic papillae or taste buds were added. Its resolution is insufficient for close inspection of actual tongue microanatomy.

Wisdom teeth, lip and cheek mucosa and frenula, intrinsic tongue muscles, styloglossus and palatoglossus, the superior alveolar and palatine nerves, the lingual artery, the pterygoid venous plexus, the maxillary sinus, minor salivary glands, all other peripheral nerves and microvessels, and complete internal tooth layers are not provided; neither Z-Anatomy nor BodyParts3D provides usable meshes for them. Some of them are drawn only as an optional schematic layer. This is not an identical reconstruction of a particular person’s mouth down to every organ and microscopic tissue.

Adapted base model: CC BY-SA 4.0. External references have separate terms. Attribution, original licenses, and modifications. Runtime notices: meshoptimizer 1.3.0 · MIT · fflate 0.8.2 · MIT. General educational anatomy; not a clinically validated model.