AI-assisted, single-fiber digital pathology quantification of changes in liver fibrosis after bariatric surgery
In brief
AI analysis detected fibrosis decline in 9 patients whose biopsy stage stayed unchanged
Among 58 patients biopsied before and after bariatric surgery, the median AI-derived fibrosis score fell from 5.25 to 3.75 over an average of nearly six years. The score tracked conventional stage changes and detected reductions in 9 patients whose stage was labeled unchanged. Whether this added sensitivity improves measurement in antifibrotic drug trials remains to be tested.
- Journal
- Hepatology communications (Q1)
- Published
- 12 February 2026
- Study design
- Unclassified
- Evidence level
- Level 5, Expert Opinion (CEBM 5)
- Authors
- Vlad Ratziu, Louis Petitjean, Raluca Pais, Judith Aron-Wisnevsky, Fréderic Charlotte, Pierre Bedossa, et al.
- PMID
- 42993745
- DOI
- 10.1097/HC9.0000000000000900
Why clinicians should know about it
- Picked for Histology (paper of the day, 9 November 2026): AI-assisted quantitative histomorphometry of liver fibrosis biopsies
Abstract
BACKGROUND: In steatohepatitis trials, antifibrotic effects are measured by conventional histology, albeit with significant limitations. We assessed how artificial intelligence-assisted, single-fiber quantitative image analysis of hepatic fibrosis matrix measures fibrosis reversal following an intervention that results in major histological improvement. METHODS: Fifty-eight patients underwent bariatric surgery with preoperative and follow-up (>1 y) liver biopsies. High-resolution, single-fiber quantitative image analysis extracted quantitative traits of fibrosis, which were combined in a continuous phenotypic fibrosis composite score (Ph-FCS) of fibrosis severity. Changes in Ph-FCS were compared between patients with increasing, decreasing, or stable histological fibrosis stage. RESULTS: The mean follow-up between biopsies was 5.84 years (SD 2.59). Forty patients improved in histological fibrosis stage, 12 were unchanged, and 6 progressed. Median Ph-FCS was reduced from 5.25 at baseline to 3.75 at follow-up (p<0.001). Baseline Ph-FCS was correlated with Ph-FCS reduction after surgery (r=0.72, p<0.001). Ph-FCS reduction was higher in patients with a histological ≥1-stage reduction (-1.78, 95% CI: -2.15 to -1.04) than in those unchanged (-1.08, 95% CI: -1.93 to 0.01) or with progression (-0.28, 95% CI: -1.26 to 0.71) (p<0.02 for differences between groups). The Ph-FCS identified intrastage fibrosis reduction in 9 patients labeled unchanged by conventional histology. Of 34 patients with advanced fibrosis at baseline, 15 still had advanced fibrosis according to histological staging. Despite similar histological staging, their follow-up Ph-FCS was reduced from baseline, although it remained higher than that of the 19 who regressed. CONCLUSIONS: Artificial intelligence-assisted, single-fiber image analysis provides an automated measurement of fibrosis changes, which is quantitative and more sensitive to change than conventional pathology and thus can improve the feasibility of antifibrotic trials.
Abstract as published, via PubMed.
For healthcare professionals. The summary is generated by AI from the published abstract, and the evidence level is assigned automatically from the study design on the Oxford CEBM hierarchy. Neither is medical advice. Read the full paper before changing practice.