Abstract
Introduction: Manual scoring of breast cancer biomarkers is subject to observer variability, and opensource digital image analysis has not been evaluated in Iraqi practice. Objective: This study aimed to compare QuPath-assisted digital scoring with manual assessment of ER, PR, HER2 and Ki-67 in invasive breast carcinomas from Salah Al-Din, Iraq, both for the individual measurements and for the surrogate molecular subtypes derived from them. Secondary aims were to characterise the cases in which the two methods disagreed, to examine how QuPath-measured HER2 membrane positivity related to the manual HER2 score, and to explore the associations of the digital readouts with standard histopathological parameters and survival. Methods: ER, PR, HER2 and Ki-67 were scored manually and with QuPath in 120 invasive breast carcinomas from Salah Al-Din, Iraq. Agreement was assessed with intraclass correlation coefficients (ICC), Lin’s concordance coefficient, Bland–Altman analysis and κ statistics, and surrogate subtypes derived by both methods were compared. Results: Agreement was excellent for all continuous measurements (ICC 0.9390.994). Categorical agreement was 99.2% for ER status, 97.5% for PR status, 97.5% for Ki-67 ≥20% and 92.5% for HER2 score (weighted κ 0.940). Subtype agreement was 114/120 (95.0%; κ 0.935). All six discordant cases had ER/PR expression of 1–10%, Ki-67 at the 20% cut-off, or HER2 upgraded from 2+ to 3+ despite non-amplified in situ hybridisation. Triplenegative status was associated with shorter disease- free survival (QuPath hazard ratio 3.65, 95% CI 1.489.00), although events were few. Conclusion: QuPath scoring agreed closely with manual assessment in this cohort. Cases near clinical cut-offs and HER2 2+ tumours still require pathologist review and in situ hybridisation. Multicentre validation is needed.