# Constructed anatomy supplements

Digestive view update (test38): the source atlas contains teeth, tongue, esophagus, uvular muscle and palatal/constrictor muscles, but no separate complete mucosal soft-palate, uvula and pharyngeal-wall surfaces. The soft palate and pendant uvula now use a Tripo-generated educational mesh, with actual surface relief, fitted to each atlas's posterior hard-palate margin and uvular-muscle landmarks. Generated arches and folds are illustrative, not donor-segmented structures. Native muscles and BodyParts3D attribution remain intact. The rejected first AI model is not included in the application. Generation provenance: https://www.magnific.com/app/creation/UPWk0V5wny . Reproducible fit: scripts/fit-generated-soft-palate.mjs and scripts/assets/soft-palate-generated.glb; audit: audit/generated-soft-palate-fit.json. Anatomy references: https://www.ncbi.nlm.nih.gov/books/NBK557817/ and https://pubmed.ncbi.nlm.nih.gov/23686451/ . These reference publications do not endorse or validate the generated mesh.

Digestive cutaway view: native jaw bones, teeth, gingiva, tongue, glands and chewing muscles use a shared front/back, left/right or top/bottom section plane confined to the head and pharynx; head/neck skin and lips are hidden. A generated posterior pharyngeal channel remains open anteriorly behind the tongue and epiglottis, then narrows below the laryngeal inlet. Its lower contour is measured by intersecting the native esophagus with a plane 4 mm below its highest vertex, rather than using the centre of the entire curved esophagus. The cutaway removes the esophageal cap at that same plane. All displayed digestive parts use the common section renderer. Only the head and pharynx, including the salivary ducts, receive the cut plane; the esophagus and abdominal organs remain whole. The shared digestive cut is left uncapped because the source contains open oral bones as well as ducts and organ shells; this prevents stencil sheets extending into empty space. Individual organ cutaways retain their own wall and cap rules. The generated wall is approximately 0.7 mm thick for display; this is not a measured mucosal thickness. Constrictor shading uses seam welding and balanced smoothing capped at 0.35 mm, without changing the source mesh. Geometry does not simulate swallowing and is not independently segmented or clinically validated. Source: https://www.ncbi.nlm.nih.gov/books/NBK54279/ . Reproducible build: scripts/add-pharyngeal-passage.mjs; audit: audit/oral-study-passage.json.

Surface-fit revision: the exposed accessory nerve display follows each model's posterolateral neck/shoulder muscle envelope. This is a presentation approximation, not an assertion that the entire nerve lies on the external trapezius surface. Regional lymph nodes with muscle overlap were locally repositioned with clearance. Numerical clearance does not establish clinical anatomical accuracy. Anatomical context: https://learning.medicine.wsu.edu/wp-content/uploads/sites/4/2023/12/PDF-Lab-29.pdf .

Lymph-node expansion: 51 selectable regional node groups per active model, replacing four broad axillary/iliac meshes with distinct subgroups and adding head, elbow, knee, thoracic and abdominal regions. Counts and positions remain illustrative. Regional references: https://anatomy.ttuhscep.edu/anatomytables/lymph_alpha.html and https://training.seer.cancer.gov/anatomy/lymphatic/components/nodes.html . No reference images or third-party node meshes were copied.

These meshes were generated locally for Human Atlas. They are educational approximations, not tissue segmentations from CT, MRI or a donor. No downloaded nerve, gland or lymph-node meshes were used. Placement was adapted to the existing BodyParts3D-based skeleton, brainstem, ribs, hand/foot bones, bladder and female presentation transforms.

Existing geometry attribution remains: BodyParts3D — The Database Center for Life Science, Kousaku Okubo and contributors — CC BY-SA 2.1 Japan. Female and Z-Anatomy source notices remain in FEMALE-REFERENCE-CREDITS.md and FACE-MODEL-CREDITS.md. These additions do not remove existing attribution or licence obligations. This notice does not certify commercial clearance of the overall app.

## What was constructed

- External spinal cord: independently measured vertebral canal landmarks in each active body, curved centreline, cervical/lower enlargement, shallow surface grooves and a tapered conus near the L1/L2 region. The pre-existing short central-canal model is retained separately.
- Representative anterior/posterior roots, 31 paired spinal nerves, dorsal root ganglia except C1, descending cauda equina roots, and an internal filum terminale reference. C1 dorsal roots/ganglia are variable and are not constructed here.
- Simplified cervical/brachial/lumbar/sacral plexuses; major nerves in the neck, trunk, limbs, palms and soles; intercostal/subcostal routes; representative cranial nerve trunks. Exact rootlets, fascicles and all terminal branches are not modeled. Plexus topology and passage through individual foramina require expert review.
- Thyroid lobes/isthmus, four parathyroid reference glands, parotid glands, palatine tonsils, representative major lymph-node groups, thoracic/right lymphatic ducts and cisterna chyli. Node counts are illustrative and variations are not reproduced.
- A short female urethral reference fitted from the current bladder outlet, anterior to the vaginal outlet reference. Pelvic fitting remains approximate.

All newly generated parts carry `Procedural educational model` provenance and a visible Thai/English notice in the anatomy panel. Shapes, routes and localization have **not** been validated by an anatomy specialist. This is a draft educational supplement, not a patient-specific model or clinical navigation tool.

## Anatomical reading used for original summaries and shape planning

- [Neuroscience — External anatomy of the spinal cord](https://www.ncbi.nlm.nih.gov/books/NBK11160/)
- [Gross quantitative measurements of spinal cord segments in humans](https://pubmed.ncbi.nlm.nih.gov/14713942/)
- [Measurements of normal cervical spinal cord on MRI](https://pubmed.ncbi.nlm.nih.gov/2107721/)
- [Neuroanatomy — Spinal nerves](https://www.ncbi.nlm.nih.gov/books/NBK542218/)
- [Texas Tech — Upper limb nerves](https://anatomy.ttuhscep.edu/anatomytables/nerves_upperlimb.html)
- [Texas Tech — Lower limb nerves](https://anatomy.ttuhscep.edu/anatomytables/nerves_lowerlimb.html)
- [Texas Tech — Lymphatics](https://anatomy.ttuhscep.edu/anatomytables/lymph_alpha.html)
- [Thyroid anatomy](https://www.ncbi.nlm.nih.gov/books/NBK470452/)
- [Clinical Methods — Neck and thyroid](https://www.ncbi.nlm.nih.gov/books/NBK244/)

No reference illustrations, tables or passages are redistributed. Summaries are newly written. Numerical dimensions are approximate planning parameters, not a statistical reconstruction of the cited subjects.

Reproducible geometry reports: models/spinal-cord-report.json, models/peripheral-nerves-report.json and models/organ-supplement-report.json. Build sources: scripts/build-spinal-cord.mjs, scripts/build-peripheral-nerves.mjs, scripts/build-organ-supplement.mjs.

Test38 pharynx: the mucosal shell is sampled from the inner surfaces of the six native constrictor muscles, with smooth interpolation across source attachment gaps and subtle modeled folds. This replaces the constant-section scaffold; the measured esophageal inlet remains unchanged. See scripts/add-pharyngeal-passage.mjs and audit/oral-study-passage.json. Abdominal membrane meshes are excluded from the shared digestive cutaway so the parietal peritoneum does not appear as an opaque enclosing sheet; they remain in their separate membrane layer. Every cutaway supports 5–100% retained depth, independent of the cut side; 100% disables surface clipping and stencil caps.

Test39 digestive study pose: lips and the lip envelope are excluded, including reference-ID variants. The mandible, lower teeth and lower gingiva share a fixed 25-degree presentation rotation around an atlas-derived jaw pivot. Tongue/floor-of-mouth and chewing-muscle copies blend toward this pose while posterior/superior attachments stay fixed. It is an illustrative viewing pose, not a dynamic TMJ or swallowing simulation. Original source geometry and the ordinary body pose are unchanged. Entry starts at 100% with the mouth open; the existing 5–100% section controls remain available. Tests: scripts/validate-digestive-half-head.mjs.


## Generated female internal reproductive study (test41)

Base: a Tripo 3D generation made for this project via Magnific, https://www.magnific.com/app/creation/mEspwoFhJQ . Original GLB is retained in scripts/assets/reproductive-generated.glb. The second generation was rejected by the provider and is not included.

The accepted mesh was edited to remove its display rod and stray strands, suppress a false raised midline vessel, close artificial ovarian holes with a subdivided envelope of the generated ovarian surface, and label 28 connected tissue regions. The anterior tract was reconstructed from the generated posterior half for the 100% state. Thin supporting folds are fitted finishing geometry. Vertex colour retains generated tissue appearance with edited uterine lining and ovarian colours. Pelvic placement is adapted to the app's native vaginal, cervical and uterine landmarks; the isolated cutaway is a straightened educational presentation. Existing anatomy identifiers describe regions; they do not imply that this generation was segmented or validated by the original atlas authors.

Anatomical layout reference: https://www.ncbi.nlm.nih.gov/books/NBK554601/ (Female Internal Genitals), and https://www.ncbi.nlm.nih.gov/books/NBK547660/ (Fallopian Tube). This is an illustrative educational reconstruction, not patient anatomy or a clinically validated surgical planning model. See audit/reproductive-generated/fit.json and validation.json for geometry checks.

## Independently generated reproductive study components (test42)

Four separate Tripo generations replace the combined test41 study source: uterus/cervix https://www.magnific.com/app/creation/vQgdXT6a47 ; vaginal wall https://www.magnific.com/app/creation/s7Gb5FOl8e ; ovary https://www.magnific.com/app/creation/rgRQFEixtc ; uterine tube/fimbriae https://www.magnific.com/app/creation/790xjxIJAL . Source GLBs are retained in scripts/assets. The final study removes extraneous structures and display fixtures, repairs a posterior lining patch using adjacent generated surface samples, mirrors the adnexa into left/right pairs, reconstructs anterior tissue for the full view, and fits supporting folds and ligaments. It is not the raw AI output or a patient scan. All original body model references remain unchanged; generated geometry appears only in cutaway mode. The 28 selectable regions are labels on these assembled source components, not 28 separately generated organs. Anatomical references above guide arrangement, not clinical certification.

## Joint adnexa and respiratory system study (test43)

The accepted adnexal source is https://www.magnific.com/app/creation/w4cxcU47EI (Tripo): ovary, tube and fimbriae generated together, fitted with shared scale and bilateral mirroring, then seating the ovary 3 mm higher into the generated fimbrial cuff. An erroneous returning proximal loop was removed and short proximal uterine connectors fitted. Earlier candidate N2dwVRr6D9 was rejected during visual review and is not used. Source w4cxZgR7EI failed at the provider. The native in-body reproductive geometry remains unchanged. Anatomical relationship reference: https://pubmed.ncbi.nlm.nih.gov/40338147/ and https://www.ncbi.nlm.nih.gov/books/NBK547660/?report=printable (ovarian fimbria and open infundibulum).

Respiratory study source: https://www.magnific.com/app/creation/SyY1RqaUb8 (Tripo). Paired lungs, trachea and airway tree generated together. A nonanatomical central ring was removed, anterior lung tissue reconstructed for full-view controls, and coarse regions labelled for atlas interaction. No claim is made that fine branch identities or regional boundaries are clinically segmented. This is an illustrative AI-derived teaching study. Original body geometry and the existing per-lung detailed study remain unchanged. The magnified alveolar viewer is a separate schematic, not actual alveoli visible at this whole-lung scale.

## Respiratory upper airway — test44
The whole-system view restores the existing native body lungs and bronchi. Separate lung cutaways keep the earlier lung study. The new upper airway is a generated 3D specimen from Tripo via Magnific (https://www.magnific.com/app/creation/VXFzhkfMMU), split into nasal cavity, pharyngeal airway and laryngeal airway with shared cut boundaries. It is fitted to native facial bone clearances and the native tracheal inlet. Facial bones, teeth and gingiva reuse existing atlas sources. The generated anatomy is an educational approximation, not a segmented or clinically validated specimen. Anatomical route: https://medlineplus.gov/ency/anatomyvideos/000059.htm. Reproductive joint adnexa are retained.

## Cranial nerve root corrections

Original SomaLens geometry corrections for both active anatomy sets: VI docks near the anterior pontomedullary junction; X docks to the lateral medullary surface; XII docks to the anterior medullary surface. Small converging rootlets accompany X and XII. XI includes five cervical rootlets (C1–C5), an ascending spinal course and an intracranial turn before rejoining the existing neck course. The 12-pair study retains those cervical roots and includes the upper spinal cord as muted context. Color, selection and model identifiers remain stable. No third-party nerve geometry was downloaded.

References: https://www.ncbi.nlm.nih.gov/sites/books/NBK544297/ and https://www.ncbi.nlm.nih.gov/books/NBK507722/ . Geometry checks confirm surface attachment and cervical bone clearance, not expert validation of exact foraminal passage, rootlet microanatomy or anatomical variation. Records: audit/cranial-root-fit/report.json.