Artificial Intelligence-Enabled Continuous HER2 Assessment: A Patient Selection Framework for the Antibody-Drug Conjugate Era.

IntroductionAntibody-drug conjugates have expanded HER2-targeted therapy to HER2-low breast cancer, yet categorical immunohistochemistry scoring inadequately supports patient selection at expression boundaries where interobserver agreement is suboptimal.MethodWe propose that artificial intelligence-enabled continuous quantification addresses this gap by preserving expression gradients essential for treatment eligibility decisions while maintaining pathologist interpretability.ResultUsing a pilot dataset (n=66), we demonstrate that data-driven thresholds derived from continuous scores can support three clinically relevant decision points: anti-HER2 therapy eligibility, in situ hybridization triage, and HER2-low identification for ADC candidacy. Bootstrap validation and cross-validation analyses supported threshold robustness.ConclusionThis transition from categorical boundaries to quantitative spectra may contribute to precision in patient selection, pending validation with treatment outcome data.