Reliable computational quantification of liver fibrosis is compromised by inherent staining variation
Bibliographic record
Abstract
Biopsy remains the gold-standard measure for staging liver disease, both to inform prognosis and to assess the response to a given treatment. Semiquantitative scores such as the Ishak fibrosis score are used for evaluation. These scores are utilised in clinical trials, with the US Food and Drug Administration mandating particular scores as inclusion criteria for participants and using the change in score as evidence of treatment efficacy. There is an urgent need for improved, quantitative assessment of liver biopsies to detect small incremental changes in liver architecture over the course of a clinical trial. Artificial intelligence (AI) methods have been proposed as a way to increase the amount of information extracted from a biopsy and to potentially remove bias introduced by manual scoring. We have trained and evaluated an AI tool for measuring the amount of scarring in sections of picrosirius red-stained liver. The AI methodology was compared with both manual scoring and widely available colour space thresholding. Four sequential sections from each case were stained on two separate occasions by two independent clinical laboratories using routine protocols to study the effect of inter- and intra-laboratory staining variation on these tools. Finally, we compared these methods to second harmonic generation (SHG) imaging, a stain-free quantitative measure of collagen. Although AI methods provided a modest improvement over simpler computer-assisted measures, staining variation both within and between laboratories had a dramatic effect on quantitation, with manual assignment of scar proportion being the most consistent. Manual assessment also most strongly correlated with collagen measured by SHG. In conclusion, results suggest that computational measures of liver scarring from stained sections are compromised by inter- and intra-laboratory staining. Stain-free quantitative measurement using SHG avoids staining-related variation and may prove more accurate in detecting small changes in scarring that may occur in therapeutic trials.
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 | 0.004 | 0.017 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| 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".