Artificial Intelligence-based Digital Pathology Assessment of CD44s Expression in Breast Cancer

Breast cancer (BC) exhibits considerable molecular and clinical heterogeneity, complicating prognostic evaluation. The cluster of differentiation 44 standard (CD44s) isoform has been proposed as a prognostic marker in various cancers; however, its role in BC remains unclear. This study evaluated CD44s expression in BC tissues and its association with clinicopathological features and survival outcomes using an artificial intelligence (AI)-based digital pathology scoring method. A retrospective analysis of 98 BC tissue samples is conducted, with CD44s cell membrane protein expression assessed through both manual and AI based immunohistochemical (IHC) scoring. Statistical analyses included Pearson’s chi-square test, Kaplan-Meier (log-rank), and Cox regression. CD44s expression was observed in 65.31% of patients. No significant associations are found between CD44s expression and clinicopathological characteristics, including age, tumor size, lymph node metastasis, histological grade, lymphovascular invasion (LVI), or hormone receptor status (all p > 0.05). Survival analysis reveals no significant association between CD44s expression and overall survival (OS, p = 0.1345) or progression-free survival (p = 0.0669). While CD44s expression is prevalent in BC samples, it is not an independent prognostic factor; LVI is the only significant predictor of OS (p = 0.036). Finally, the moderate agreement between AI and manual scoring (Cohen’s Kappa = 0.4337, p < 0.0001) supports the potential of AI-assisted methods for biomarker quantification, warranting further validation in larger cohorts.