A mouse cerebral infarct severity scoring system based on image color difference algorithm

Ischemic brain damage in transient middle cerebral artery occlusion (tMCAO) animal model is commonly assessed by measuring infarct volume based on 2,3,5-triphenyltetrazolium chloride (TTC) staining of brain sections. Analysis of TTC stained brain slices traditionally depends on manual delineation of the boundary of infarct regions, which introduce subjectivity and is labor-intensive. Previous approaches for automated analysis often lack precision in accurately identifying infarct boundaries. Furthermore, the ischemic penumbra (IP) is consistently overlooked in both manual and automated analyses, despite its critical role in therapeutic salvage and stroke prognosis. Following a neurological behavioral assessment, TTC-stained slice images were acquired 24 h post-tMCAO surgery and subsequently analyzed using manual measurement, automated measurement, and our proposed severity scoring system. For manual measurement, experienced observers manually delineated the infarct area without distinguishing IP and ischemic core (IC). For automated measurement, image segmentation was performed in Fiji using built-in auto-thresholding algorithm. In our scoring system, images were first decomposed into split RGB and HSV channels. Subsequently, the pixel value histograms of each channel were analyzed to label the IP and IC in the images, and then generated quantitative scores via our algorithm based on the volume of IP and IC and the mean pixel value of IP. These scores were compared via Pearson correlation with the manual measurement to identify the optimal metric. Spearman correlation analysis between the score and the neurological behavioral assessment was conducted to evaluate the correlation between the scoring system and overall cerebral motor function. Additionally, photothrombotic (PT) ischemic stroke model was utilized to validate the generalizability of our proposed severity scoring system. The thresholds for distinguishing the IP and IC were established for each channel. The blue channel value showed the strongest correlation with manual measurement and exhibited excellent recognition of the IP and IC, thus identifying it as the optimal metric for the algorithm. The correlation coefficients between the scoring system’s output (i.e., Severity ) and either manual or automated measurement were 0.9392 and 0.8885, respectively. Severity demonstrated a strong correlation with the neurological deficit score (Spearman r = 0.7492), outperforming both manual (Spearman r = 0.7206) and automated measurements (Spearman r = 0.6528). Furthermore, Severity also showed a significant correlation with manual measurement in the PT ischemic stroke model. Our scoring system successfully distinguishes the IP and IC regions. Behavioral validation confirms that this scoring system reliably assesses infarct severity and thus accurately reflecting the overall severity of cerebral infarction induced by tMCAO. Moreover, the scoring system is also applicable to the PT stroke model. Taken together, this severity scoring system represents a novel method for TTC staining analysis. It is expected to provide a rapid, objective, and accurate tool for the high-throughput evaluation of stroke outcomes in basic research.