Letter to the Editor “Artificial Intelligence in digital pathology of breast cancer, new era of practice?”

Dear Editor, We read with great interest the recent article titled “Artificial Intelligence in digital pathology of breast cancer, new era of practice?”[1], which provides a systematic review of the latest research advances in artificial intelligence (AI) in digital pathology of breast cancer, focusing on its applications in key areas such as the recognition of histological features, the evaluation of biomarkers, and the integration with multi-omics data. We greatly appreciate the authors’ timely and comprehensive review on this challenging topic. However, we believe that certain limitations remain in the current review, which may affect its scientific rigor and practical value in guiding clinical applications. First, the review lacks systematic evaluation of the methodological quality of the included studies, such as sample size, risk of bias, and external validation robustness – within an established evidence-based framework like standardized systematic review guidelines[2]. Notably, a considerable proportion of the cited studies rely on small, single-center retrospective datasets, which may introduce sample-selection bias and increase the risk of overfitting. Second, while the article acknowledges the challenges facing AI in clinical application, it lacks an in-depth discussion grounded in surgical practice. Specifically, it does not adequately explore how AI models can effectively integrate into critical intraoperative decision-making processes, such as rapid frozen section diagnosis and real-time tumor margin assessment. Additionally, several pivotal translational and regulatory issues remain unaddressed, including the integration of AI models with existing hospital information systems, pathways for regulatory approval, mechanisms for insurance reimbursement, and the delineation of legal liability and ethical responsibility in AI-assisted diagnostics[3–5]. Third, the performance of AI models is highly dependent on the quality of training data[5]. This review does not systematically address the challenges associated with acquiring high-quality whole slide images, nor does it adequately discuss the additional difficulties posed by the specific characteristics of breast cancer pathology samples – such as small biopsy tissues or the heterogeneity in tumor regression samples following neoadjuvant therapy – for AI model training. Furthermore, insufficient emphasis is placed on the challenges of constructing large-scale datasets that require multidisciplinary annotation involving teams from surgery, pathology, and radiology, which is essential for achieving “surgical precision medicine.” In conclusion, the field of AI-driven digital pathology demonstrates immense potential and represents a critical step toward precision pathology, yet it continues to face dual challenges in both technical implementation and clinical integration. We commend the authors for their valuable contribution, as this review serves as an authoritative academic reference for researchers in related fields. By incorporating analyses that are more critical and closely aligned with the practical needs of surgical pathology, the scientific value of this review and its role in advancing the discipline will be significantly enhanced.