Digital image analysis improves diagnostic accuracy of HER2-low and HER2-ultralow breast cancer: a step towards personalised medicine.

Accurate assessment of human epidermal growth factor receptor 2 (HER2)-low and HER2-ultralow breast cancer is challenging, and digital image analysis (DIA) may provide greater reproducibility. This study aimed to evaluate the concordance between DIA and expert pathologist consensus for HER2 status assessment in breast cancer, including HER2-low status. Furthermore, differences in staining intensity between HER2-ultralow and HER2-null status were investigated. Additionally, concordance between initial pathologist assessment and expert pathologist consensus was evaluated. HER2 immunohistochemistry slides from 102 consecutive retrospective breast cancer biopsies were scanned. Invasive carcinoma regions were annotated, and DIA was performed to quantify HER2 membrane staining intensity. An initial HER2 score assigned by the routine pathologist at the time of consensus was collected for comparison. An adjudication consensus diagnosis established by three expert breast pathologists was used as the ground truth. Concordance was assessed using Cohen's weighted κ. DIA correctly identified 84% of cases, with concordance between DIA and pathologist consensus found to be almost perfect (weighted κ = 0.84). DIA demonstrated higher concordance with the ground truth than the initial pathologist assessment (weighted κ = 0.76). HER2-ultralow cases showed significantly more weakly membrane-stained cells than HER2-null cases (p < 0.0001). A threshold of 6.13% stained cells differentiated between the two statuses with 81% accuracy and 100% specificity. In conclusion, DIA is a reliable and objective method for HER2 status assessment, particularly in distinguishing between HER2-low and HER2-ultralow tumours, with the potential to improve patient stratification and treatment selection.