Repeatability and Reproducibility of Artificial Imaging-based Digital Pathology for the Evaluation of Liver Fibrosis

Fibrosis stage is a key determinant of outcomes in metabolic dysfunction-associated steatohepatitis (MASH). Assessment of fibrosis change is a critical component of the assessment of therapeutic efficacy of interventions. Conventional assessment of fibrosis stage is limited by sampling and inter- and intra-observer variability. Second harmonic generation/2-photon excitation fluorescence (SHG/TPEF) imaging provides an alternate way to assess fibrosis from unstained histologic sections. The aim of this study is to establish repeatability and reproducibility of SHG/TPEF imaging in metabolic dysfunction-associated steatohepatitis histologic samples.

The SHG/TPEF images were acquired by 3 Genesis machines for the included samples. Each sample was scanned 3 times by each machine at different time points. qFibrosis stage (qFS) and qFibrosis continuous value (qFC) were calculated by an artificial intelligence-based algorithm that was previously correlated with the Nonalcoholic Steatohepatitis Clinical Research Network fibrosis staging system.

The overall intra-/inter-system weighted kappas of qFS were 0.880 (95% confidence interval [CI], 0.844-0.915) and 0.850 (95% CI, 0.811-0.889), respectively. Using the pairwise agreement method, the intra-/inter-system agreements of qFS were 89.16% (95% CI, 0.856-0.921) and 86.46% (95% CI, 0.829-0.890), respectively. For qFC, the intra-/inter-system agreements were 87.53% (95% CI, 0.837-0.911) and 78.05% (95% CI, 0.737-0.824), respectively. Furthermore, the Bland-Altman method was used to evaluate the intra-/inter-system agreements of qFC, with 93.77% (95% CI, 0.911-0.960) and 93.50% (95% CI, 0.908-0.959), respectively.

The qFibrosis system has good repeatability and reproducibility for fibrosis staging. It can provide an accurate reference for pathologists to stage liver fibrosis more objectively.