Skip to content

Histopathology finding_term NCIT triage

Provenance note for the HistopathologyFindingTerm branch expansion and the subsequent backfill of histopathology[].finding_term bindings.

Background

HistopathologyFindingTerm was widened from Morphologic Finding (NCIT:C35867) + Histologic Grade (NCIT:C18000) to cover the whole NCIT Histopathology Result branch (NCIT:C83490), adding Immunophenotypic Finding (NCIT:C40998), Ultrastructural Finding (NCIT:C43265), and Staining Intensity (NCIT:C127762). See CLAUDE.md and the schema (src/dismech/schema/dismech.yaml).

Coverage at time of triage

At the start there were 436 histopathology findings, of which 282 carried a finding_term. This triage bound 31 more across three batches:

  • Neoplastic/heme batch (11): Carney-Stratakis (SDHB-loss IHC, epithelioid component, lymphatic invasion), Anal_Canal_Adenocarcinoma (mucinous adenocarcinoma, cytokeratin-positive immunophenotype), Penile_Cancer & Squamous_Cell_Carcinoma_of_Penis (squamous cell carcinoma), Myeloproliferative_Neoplasm_Unclassifiable (megakaryocyte proliferation), Aggressive_NK-cell_Leukemia (bone marrow involvement), Meningeal_Melanocytoma (meningeal melanocytoma, HMB45-positive marker panel).
  • Non-neoplastic generic-pathology batch (20): inflammatory/lymphoplasmacytic/ eosinophilic infiltrate, fibrosis, necrosis, acantholysis, hyperkeratosis, spongiosis, hypercellularity, nuclear inclusion, amyloid/hemosiderin deposition, villous atrophy, metaplasia, onion-skinning — see the per-file finding_term blocks.

Key finding: where NCIT does and does not reach

The NCIT Morphologic Finding branch has good coverage for generic pathology patterns (fibrosis, necrosis, granulomatosis, metaplasia, hyperplasia, inflammatory/eosinophilic/lymphoplasmacytic infiltrate, amyloid/hemosiderin deposition, acantholysis, spongiosis, hyperkeratosis, villous atrophy, nuclear inclusion) and, via the newly added Immunophenotypic branch, for IHC markers (SDHB loss, cytokeratin, HMB45, ER/PR/HER2, Ki-67 percentage bins).

It is sparse or absent for:

  1. Organ-specific microscopic findings, especially renal (foot-process effacement, mesangial hypercellularity as a named term, glomerular crescents, segmental glomerulosclerosis, tip lesion) — not in the branch.
  2. A plain "Granuloma" / "granulomatous inflammation" morphologic finding. NCIT:C3064 (Granuloma) is a Disorder node, not reachable from the Histopathology Result branch; the in-branch granuloma terms are all tumor/entity-specific (Lymphomatoid Granulomatosis, Xanthogranuloma, etc.). Neurosarcoidosis and Farber disease granuloma findings were therefore left unbound rather than mis-bound.
  3. Disease/eponym-level "findings" that are really entities, not morphology (Barrett esophagus, Castleman hyaline-vascular/plasma-cell variants, DNET "specific glioneuronal element", Touton giant cells, juvenile polyp).
  4. HP is not a substitute here. finding_term currently allows only HP:0025461 (Abnormal cell morphology) descendants — 60 terms, almost all cell-level (rosettes, inclusion bodies, Gaucher cells). It does not cover tissue/architectural findings, so the ~123 remaining unbound findings (mostly non-neoplastic renal/derm/neuro/muscle/metabolic) cannot be rescued by the current HP root.

Remaining unbound: 123 findings across 76 files

These fall into three buckets:

  • No faithful morphologic-finding term in NCIT and outside the current HP root (renal glomerular findings, demyelination patterns, muscle-biopsy findings such as ragged-red fibers / nemaline rods, corneal deposits, disease-specific patterns). Best left as prose unless the schema HP root is broadened.
  • Bindable only to an over-generic NCIT term (e.g. everything → "Deposit") where the generic term adds little machine-queryable value; deferred.
  • Genuinely entity-level ("findings" that are diagnoses); these arguably belong in disease_term/subtype rather than histopathology.

Recommendations (open)

  1. Consider broadening the HP root for HistopathologyFindingTerm beyond HP:0025461 to a histology-relevant set (e.g. specific organ-histology HP branches) so non-neoplastic tissue findings become bindable — this is the single biggest lever for the remaining 123. Needs a design-decision entry because it risks blurring the histopathology/phenotype boundary.
  2. For the renal findings specifically, evaluate whether an additional ontology (e.g. a renal-pathology vocabulary) is warranted, or accept prose.
  3. Leave entity-level "findings" as prose or migrate them to disease_term.