Abstract / Summary
Respiratory viral infections such as influenza and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) remain major causes of morbidity and mortality. Although lung histopathology is a major endpoint in preclinical studies, conventional H&E-based grading primarily evaluates the extent of inflammatory-cell infiltration and may not distinguish progressive from resolving disease. Here, we investigated whether inflammatory-cell composition could refine lung histopathology assessment using convolutional neural network (CNN)-based analysis. In SARS-CoV-2 mouse cohorts, systemic disease severity was classified according to body-weight loss and splenic white-pulp atrophy. H&E-based inflammatory-cell infiltration scores and CNN-derived inflammatory-cell-rich areas did not fully correspond to systemic disease status in several infection and vaccine-challenge settings. In contrast, lesions associated with lower disease-severity or recovery-phase states were enriched in lymphocytes and CD8⁺ T cells, whereas higher disease-severity or lethal-phase lesions showed macrophage- and neutrophil-rich patterns. A two-step CNN workflow that detected inflammatory-cell-rich regions and subsequently classified them as recovery-like or lethal-like supported this composition-based distinction across independent SARS-CoV-2 cohorts. Similar lymphocyte- and CD8⁺ T-cell–enriched inflammatory patterns were also observed during recovery in influenza A infection. These findings support lymphocyte predominance as a practical histopathologic parameter that complements conventional inflammatory-cell infiltration scoring in murine respiratory viral infection models.