Wide-Angle Lung Experiment Segmentation (WALES): Effective quantitative assessment of lung pathology in model systems.

Abstract

For preclinical studies, the standard practice for evaluating lung injury usually involves an assessment of pulmonary histopathology by a certified pathologist. This is typically accomplished by light microscopy using a semiquantitative four-point scale. In contrast, automated image analysis software allows a more quantitative assessment, though inherent limitations with such automated programs can produce misleading conclusions. For example, specific imaging features may be incorrectly scored or classified within the specimen because of the complex architecture and heterogeneous structures present in the lung. In addition, tissue processing and handling may further introduce artifacts and inconsistencies that affect automated analysis. To address these limitations, we developed a lung image analysis program, Wide-Angle Lung Experiment Segmentation (WALES), which uses Meta's Segment Anything Model to provide semiautomated masking and relative density analysis to efficiently quantify lung injury. Density analysis using WALES effectively delineated varying severities of lung injury, not achieved using more standard methods. WALES is widely applicable for many preclinical lung injury models.