Artificial intelligence in breast cancer diagnosis through histopathology and biomarker detection: a scoping review

Breast cancer diagnostics depend critically on the accurate histopathological evaluation and identification of tumour biomarkers in biopsies and serum. While standard operating procedures (SOPs) provide structured diagnostic workflows, they are constrained by inter-observer variability, staining inconsistencies, and inconsistencies in interpretive criteria and decision thresholds. Artificial intelligence (AI) offers promising solutions by integrating heterogeneous data sources and standardising interpretation. This scoping review critically evaluates 11 recent studies on AI-driven approaches for tumour biomarker detection using serum panels, immunohistochemistry (IHC) images, digital histopathology, and morphology-based prediction. AI techniques such as convolutional neural networks, gradient-boosted machines, and ensemble models, have demonstrated improvements in diagnostic performance, reproducibility, and efficiency, achieving up to 92.9% specificity in serum-based models and image-based classifications. Several studies have shown that AI can infer molecular biomarker status directly from H&E-stained slides, thereby eliminating the need for biochemical assays. Despite these advances, significant challenges persist. Many models lack external validation, explainability, and integration into clinical workflows. Furthermore, concerns regarding data bias, ownership, and equitable deployment highlight the urgent need for ethical governance and inclusive model design. Future research should prioritise prospective validation, multimodal data integration, and clinician-centred implementation strategies to realise AI’s potential in advancing breast cancer diagnostics and personalised care.