Interstitial lung diseases (ILDs) comprise a heterogeneous group of pulmonary disorders marked by chronic inflammation and/or fibrosis. The authors note that 30–40% of ILD patients progress to fibrotic disease with progressive respiratory decline and poor prognosis, particularly in idiopathic pulmonary fibrosis. Current antifibrotic therapies can slow disease progression but do not reverse established fibrosis, highlighting an ongoing need for improved therapeutic strategies and objective preclinical readouts.
Conventional histopathological evaluation in preclinical ILD models relies mainly on semi-quantitative scoring systems that are time-consuming, subject to inter-observer variability, and limited by restricted field sampling. To address these limitations, the authors developed FibroSight, a standalone, user-oriented platform for automated, multi-compartment histological quantification in Sirius Red–stained lung sections.
FibroSight combines deep learning–based structural segmentation with color-based feature extraction to enable largely automated whole-lobe analysis. The platform is designed to run without a complex computational setup and to perform compartment-resolved segmentation of lung tissue, distinguishing relevant structures to support targeted quantification.
By integrating segmentation outputs with colorimetric analysis of Sirius Red staining, FibroSight computes features that reflect collagen deposition and tissue remodeling across parenchymal, airway, and vascular compartments. The authors describe the platform as standalone and scalable, intended to reduce manual labor and inter-observer variability inherent to traditional scoring approaches.
FibroSight produces a set of complementary remodeling readouts to capture multiple aspects of lung pathology. Key metrics reported by the platform include:
These metrics are intended to provide a broader, multi-dimensional view of remodeling than single-score approaches, enabling separation of fibrotic, inflammatory, airway, and vascular contributions to tissue pathology.
The authors validated FibroSight in the commonly used bleomycin-induced fibrosis model. FibroSight-derived quantitative metrics correlated strongly with expert Ashcroft scoring, a standard histological assessment for fibrosis severity. When compared with a semi-automated ImageJ–based workflow, FibroSight outputs demonstrated stronger associations with histological severity, indicating improved sensitivity or fidelity of the platform’s multi-compartment measures in this preclinical model.
Details of the validation cohorts, sample sizes, or specific correlation coefficients were not reported in the source summary and therefore are not provided here.
Beyond fibrosis quantification, FibroSight was applied to an influenza-induced lung injury model where it differentiated inflammatory remodeling from fibrotic remodeling. This indicates the platform’s capacity to resolve distinct pathological processes within stained sections and suggests utility for studies where distinguishing inflammation-driven changes from collagen deposition is important.
The source text does not provide granular methodological details on how inflammatory features were discriminated, nor does it report numerical performance metrics for this comparison.
To demonstrate translational potential, the authors applied FibroSight to human ILD biopsy specimens as a proof of concept. This application suggests the platform can be extended from preclinical whole-lobe analyses to human diagnostic material, supporting cross-species or translational studies of lung remodeling.
Specific results, sample characteristics, or performance metrics for the human specimens were not detailed in the source summary and thus are not reproduced here.
The authors position FibroSight as addressing key limitations of current histopathological workflows by offering:
The platform’s stronger associations with histological severity compared with a semi-automated ImageJ workflow were highlighted as evidence of improved quantitative performance in the bleomycin model.
The authors report that code and resources for FibroSight are available at the project repository: https://github.com/Anas-Odeh/FibroSight. The preprint states that the authors have declared no competing interests.
FibroSight provides an AI-assisted framework for objective, multi-compartment histological quantification in Sirius Red–stained lung sections. By combining deep learning segmentation with color-based feature extraction, the platform yields multiple complementary metrics — including parenchymal collagen fraction and other compartment-specific readouts — and enables automated whole-lobe analysis without complex computational requirements. Validated in the bleomycin model and demonstrated in influenza injury and human biopsy specimens, FibroSight aims to support more precise analysis of disease mechanisms and therapeutic responses in preclinical and translational ILD research.
Further methodological details, quantitative performance metrics, and sample-level data are available in the full preprint and the authors’ referenced repository, as the source summary does not report those specifics.