Abstract 34: Predictors of discordance between image-based and manual scoring of immunohistochemical stains in ovarian cancer tissue microarrays
Bibliographic record
Abstract
Abstract Digital image-based scoring of immunohistochemistry (IHC) staining allows faster, more objective and potentially more reproducible quantification of area stained or cell counts compared to manual scoring. However, predictors of discrepancies between these two assessment methods are not well understood. We examined clinical and pathological characteristics associated with discordance between image-based and manual scoring using ovarian cancer tissue microarrays (TMA) across a range of IHC stains. We evaluated 2,159 TMA cores from 681 ovarian cancer cases stained by four tissue cytoplasmic stains quantified as percentage of the tumor area stained (POSTN, CXCL14, ADH1B, COL11A1) and two immune cell markers quantified by cell count of infiltrating immune cells (CD68, CD163). The TMA slides were scored manually using semi-quantitative scoring by gynecologic pathologists as well as digitally using image-based Definiens automated platform. We used generalized linear mixed models with individual patients and TMAs included as random effects and stain as fixed effect to examine predictors of discordance, defined as absolute difference of ≥1SD in z-scores between image-based and manual scoring, overall and by each stain. The multi-level model included characteristics of the TMA, core, and tumor factors as predictors. Overall, TMA construction by automated robotic system (vs manual), smaller core size or area, borderline (vs invasive tumors), and having no geographic necrosis in the tumor were significantly associated with greater discordance between image-based and manual scoring. Interestingly, all these factors, except geographic necrosis, were differentially associated with discordance across the stains (p-interaction <0.05). Having multiple loci of geographic necrosis was associated with less discordance (OR=0.84, 95%CI=0.72-0.97). Invasive tumors compared to borderline tumors were less likely to be discordantly scored overall (OR=0.80, 95%CI=0.65-1.00) and for tissue cytoplasmic stains (OR range: 0.42-0.73), with associations in the opposite direction for immune cell markers (OR range: 1.21-2.37). Core areas in the lowest quartile vs all other cores were associated with greater discordance (OR=1.18, 95%CI=1.05-1.32), which was more notable for the tissue cytoplasmic stains (OR range: 1.18-2.00) compared to immune cell markers (OR range: 0.83-0.86). TMAs created by robot vs by hand had 64% higher odds of discordance (95%CI=1.20-2.25), which was more apparent when blank rows and columns were included on the TMA. Overall, TMA, core, and tumor level factors were related to discordance between image-based and manual scoring. Some stain types may be more susceptible to specific pre-analytic factors suggesting that reproducibility studies of manual vs image-based scoring should be conducted on a proportion of cases in large scale projects. Citation Format: Naoko Sasamoto, Mary Townsend, Farnoosh Abbas-Aghababazadeh, Kathryn L. Terry, Joseph O. Johnson, Jonathan L. Hecht, Brooke L. Fridley, Shelley S. Tworoger. Predictors of discordance between image-based and manual scoring of immunohistochemical stains in ovarian cancer tissue microarrays [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 34.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.016 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".