Background: Traumatic brain injury (TBI) induces spatially heterogeneous neuronal injury that extends beyond the acute phase, yet robust and scalable methods for identifying brain-wide pathological patterns remain limited. Traditional histological techniques, such as silver staining, provide high-resolution structural information but are limited in their ability to quantify spatial injury patterns across the entire brain. Here, we developed a data-driven computational framework that integrates high-throughput histology with supervised machine learning to characterize spatial patterns of neuronal injury following repetitive TBI in mice. Methods: Adult male C57BL/6 mice underwent single or repetitive (two-impact) midline fluid percussion injury, and brain tissue was collected post injury for histological analysis. Using the Amino CuAg (de Olmos) silver staining method, we quantified neuronal injury across whole coronal hemisections spanning the anterior–posterior axis and extracted pixel-level intensity and anatomical features. These features were used to train a Random Forest classifier to predict injury severity based on silver staining-derived metrics, with model performance evaluated using receiver operating characteristic analysis. Results: The classifier achieved high predictive accuracy (area under the curve = 0.90), successfully distinguishing sham, single-, and repetitive-injury groups. Model predictions aligned with spatial patterns of silver staining, with peak injury localized near the bregma–lambda epicenter and extending into posterior cortical regions across time points. The model assigned progressively higher predicted injury severity scores across sham, single-, and repetitive-injury samples, demonstrating clear separation among injury conditions. Conclusions: Together, these findings establish a scalable computational workflow that links high-throughput histological imaging with predictive modeling to quantify brain-wide neuronal injury. This approach provides a foundation for developing objective and reproducible biomarkers of injury burden following TBI.
Lugo et al. (Sun,) studied this question.