Editor's Note: Pathological diagnosis is a cornerstone of precision breast cancer care, and standardization and digital and intelligent transformation are key issues in the ongoing development of pathology. At the 2026 CSCO Annual Meeting, Oncology Frontier invited Professor Hong Bu of West China Hospital, Sichuan University, to share his insights on standardizing breast pathology diagnosis, approaches to digital and intelligent transformation, and collaboration between pathology and clinical departments.Oncology Frontier: Standardization of breast pathology diagnosis has continued to advance in recent years, from increasingly refined HER2 testing procedures to the expansion of molecular pathology testing. Pathology reports are now expected to contain increasingly comprehensive information. From the perspective of internal development within the pathology discipline, which aspects of standardization in breast pathology diagnosis most urgently need improvement? What are the main challenges faced by medical institutions at different levels when implementing standardized practices?
Professor Hong Bu: I believe that pathological diagnoses relevant to the clinical management of breast cancer can broadly be divided into two major areas: histopathological classification and molecular classification. Molecular classification is more closely associated with the reproducibility of immunohistochemical assessments. We have done extensive work in this area and discussed many related issues. However, reproducibility in histopathological diagnosis is the aspect of clinical practice that requires greater emphasis on accumulated experience, continuous refinement, and ongoing improvement over the long term.
I divide this issue into three parts. First, we need to know the standards. This is extremely important, and many consensus frameworks are already available to help us establish clear standards. Second, we need to understand the standards. There are considerable differences in how individuals interpret them, and this is one reason why pathological diagnoses may lack reproducibility. Even when applying the same standard, different people may understand and interpret it differently. For example, everyone recognizes the importance of tumor atypia, but some may consider the atypia sufficient, whereas others may feel that it does not quite meet the required threshold. This is precisely where we need to continue learning, deepening our understanding, and improving our diagnostic capabilities. Many educational lectures today are essentially designed to help pathologists develop a better understanding of the standards.
The third step is to apply the standards, which can also be understood as knowing when to go beyond them. At this stage, we need to make greater use of clinical information to guide our judgments. For example, factors such as the patient’s age and tumor size may help us assess findings that have not yet been definitively standardized but are highly meaningful in clinical practice. These considerations are also important for arriving at the final diagnosis. If we can master all three steps, the reproducibility of histopathological diagnosis will improve, and diagnostic accuracy will increase as well.
Oncology Frontier: In recent years, the application of digital scanning and artificial intelligence (AI) in pathology has continued to expand, with breast pathology being one of the areas making relatively rapid progress. In routine pathology department operations, digital and intelligent transformation involves multiple aspects, including hardware infrastructure, data storage, staff training, and workflow adjustments. What foundational requirements should be prioritized in the current digital and intelligent transformation of breast pathology? Should implementation pathways be designed differently for medical institutions at different levels?
Professor Hong Bu: I believe that large hospitals and primary-level hospitals will inevitably need to follow different pathways in the digital and intelligent transformation of breast pathology. The key is to take each institution’s actual circumstances into account and identify the specific problems and pain points in its pathological diagnostic practice. AI should be used to address clinical pain points. Wherever those pain points lie, AI should be deployed to help resolve them. Consequently, the approaches to digital and intelligent transformation will inevitably differ from one institution to another.
For breast pathology as a subspecialty, we need to consider how AI is currently developing and apply it accordingly. Broadly speaking, AI has progressed through several stages. The first stage is digitization. The second stage involves AI-assisted tasks such as quantification and grading. We are now also exploring the use of AI to predict patient outcomes and guide treatment. These applications go beyond the capabilities of traditional human assessment and offer considerable room for further development. However, the available data have not yet been sufficiently validated, and this remains an important direction for future efforts. The third stage involves integration: bringing together clinical, imaging, and pathological information to achieve more accurate results than diagnosis based on any single information source alone.
At present, I believe the second stage is more readily implementable. It can help improve the reproducibility of assessments that have traditionally relied on qualitative and quantitative judgments. Ki-67 assessment is one example. Clinicians need accurate results in their reports, but in actual practice, the reproducibility of these assessments still needs improvement. AI can substantially help us address problems of this kind.
The storage of breast pathology specimens is currently a major obstacle. To address this issue, I recommend that many institutions begin with biopsy specimens. There are several reasons for this. First, the number of pathological biopsy specimens obtained from patients is increasing. Second, many diagnostic pain points associated with biopsy specimens can be addressed using AI. Third, biopsy specimens require relatively little storage space. In addition, pathologists are generally more accustomed to making digital diagnoses on a screen when working with small specimens. This gives us a practical and feasible starting point for implementation.
Oncology Frontier: The ultimate value of pathological diagnosis lies in its ability to support clinical treatment decisions. Looking at the overall development of breast pathology, which aspects of collaboration between pathology departments and clinical departments most urgently need to be strengthened? As diagnostic and treatment approaches continue to evolve, how are communication and feedback mechanisms between pathology and clinical teams changing?
Professor Hong Bu: This is one of the issues we care about most in clinical practice. The CSCO established its Expert Committee on Tumor Pathology precisely to promote closer integration between pathology departments and clinical departments.
For more than a decade, one of the principles I have most strongly advocated is that pathological diagnosis should be based on the individual patient’s clinical circumstances. In other words, pathological diagnosis must meet clinical needs and serve clinical practice. It should not simply involve pathologists developing increasingly detailed and complicated histopathological classifications for their own sake. Every classification system and diagnostic method must have clinical significance, while reproducibility must also continue to improve. These are principles I have consistently worked to advance.
Of course, we must also maintain close communication between pathology and clinical teams, helping clinicians acquire a deeper understanding of pathology while enabling pathologists to better understand clinical treatment. This is work we have continued to pursue. We are striving to promote education in both directions: first, ensuring that clinicians understand the pathological knowledge they need; and second, ensuring that pathologists acquire the clinical knowledge required for their work. This is an important aspect of our efforts and one of our ongoing goals.
Professor Hong Bu
