Quantitative and unbiased lung alveolar septum assessment in a LPS experimental mouse model using 2D-spatial correlation image analysis from hematoxylin and eosin slides

Quantitative assessment of lung tissue architecture is essential for evaluating disease progression in experimental models of acute lung injury (ALI). However, conventional methods for measuring alveolar septum (tabique) thickness rely on manual annotation, which is time-consuming and observer-dependent, compromising comparability and reproducibility. In this study, we introduce a novel, fully automated approach based on two-dimensional spatial autocorrelation function (2D-ACF) analysis to quantify septum thickness from hematoxylin and eosin (H&E)-stained lung tissue sections.

After validating the method with a simulated data set, the new approach was applied to a murine model of Acute Lung Injury (ALI) induced by intratracheal instillation of lipopolysaccharide (LPS), under two dietary conditions: normal (ND) and high-fat diet (HFD), thereby testing our approach with a double hit protocol for ALI development. The 2D-ACF analysis provided a robust metric of structural organization, allowing for an unbiased estimation of mean septum thickness across entire tissue images. Compared to controls, LPS instillation increased septal thickness more than six-fold, with further thickening observed in the HFD+Instilled group. These findings were consistent across >400 images and captured subtle additive effects of metabolic stress on lung injury.

Traditional manual measurements of septum thickness exhibited substantial inter-observer variability, particularly in the LPS-instilled group, where structural heterogeneity made consistent interpretation challenging. This subjectivity limits the reproducibility of histopathological evaluations, especially under pathological conditions with subtle anatomopathological differences. In contrast, the 2D-ACF method offered a standardized and observer-independent approach. A direct comparison between methods revealed a strong and statistically significant correlation in the instilled group (r = 0.89, p < 0.001), indicating that the 2D-ACF captures key structural features aligned with expert assessments while reducing user bias.

In summary, the 2D-ACF framework offers a powerful alternative to conventional image analysis for studying lung pathology. Its adaptability to different experimental settings and potential to be extended to other tissues with spatial heterogeneity makes it a valuable tool for translational biomedical research and digital pathology.