Editor's Note: Pathological diagnosis is the cornerstone of precision breast cancer care. In recent years, rapid advances in digital scanning, artificial intelligence (AI), and multimodal integration have driven a profound transformation in breast pathology, moving it beyond traditional morphological diagnosis toward a digital and intelligent paradigm. These developments offer new tools and evidence to support precise breast cancer classification, treatment response prediction, and individualized therapy. The 17th Yanzhao Breast Cancer Forum was held in Shijiazhuang from August 26 to 30, 2026. During the meeting, Oncology Frontier invited Professor Hong Bu of West China Hospital, Sichuan University, to discuss the development of digital and intelligent breast pathology, drawing on his presentation entitled “The Development of Digital and Intelligent Breast Pathology.” He shared his insights into the key stages of development, the clinical implementation of AI and human–AI collaboration, the digital transformation of pathology, and its clinical value.01
Oncology Frontier: At this year’s Yanzhao Breast Cancer Forum, you delivered an important presentation entitled “The Development of Digital and Intelligent Breast Pathology,” offering guidance on the future development of breast pathology in this new era. Could you outline the key milestones in the development of digital and intelligent breast pathology in China?
Professor Hong Bu: The development of digital and intelligent breast pathology can be broadly divided into four stages.
The first stage is digitization, which provides the foundation for the entire field. The second involves quantitative analysis and improvements in diagnostic reproducibility based on digital pathology. The third uses digital pathology and AI technologies to predict treatment efficacy, prognosis, and therapeutic response. The fourth is the integration of multimodal data, including clinical information, imaging data, and pathological knowledge.
Of these four stages, digitization remains the most critical and urgent priority. Many applications in digital and intelligent pathology have yet to achieve widespread clinical implementation because the necessary digital infrastructure is not yet sufficiently established.
02
Oncology Frontier: How is AI implementation progressing in breast cancer pathology? What is your response to concerns among some physicians that AI may replace pathologists?
Professor Hong Bu: Early adopters are already implementing AI in breast cancer pathology, and there are signs that its use is beginning to expand. The key to implementation is the scale of digitalization within pathology departments, although some departments can also adopt a “start small and move quickly” strategy.
One practical starting point is breast biopsy specimens. These generally place a relatively modest burden on storage capacity, while clinical practice often requires timely reporting. This makes them a suitable entry point for initial implementation. Overall, however, further efforts are needed before AI can be deployed comprehensively in real-world practice.
The idea that “AI will replace pathologists” is, in my view, a misunderstanding. The future of pathology will be based on a scientific and harmonious model of human–AI collaboration. From both ethical and scientific perspectives, the role of human professionals remains irreplaceable. Pathologists’ workflows will certainly change: their workloads may decrease, and the quality of their work may improve. However, this does not mean that pathologists will no longer be needed.
Nevertheless, pathologists today must learn about AI, understand how it works, and know how to use it. Those with strong AI literacy will be more competitive in future clinical practice and will have greater opportunities for professional development.
03
Oncology Frontier: From the perspective of pathology departments, what barriers must be overcome in terms of hardware infrastructure and changes in mindset to achieve genuine digital and intelligent transformation?
Professor Hong Bu: From a hardware perspective, digital scanners are no longer the main obstacle. As equipment prices decline and the National Healthcare Security Administration advances relevant reimbursement and pricing policies, digital scanning is expected to become more widely available in the near future.
Recently, the National Healthcare Security Administration released a proposal to establish pathology cloud platforms. It proposes assigning each patient who undergoes pathological examination a unique identifier, similar to a national identification number, so that pathological slides can be traced and shared across institutions through the cloud platform. This could reduce the need to physically transfer slides between institutions in the future, although implementation will take time and must proceed gradually.
The main challenge at present is data storage. Scanning pathological slides generates enormous volumes of data, and institutions that have begun exploring digitalization commonly face substantial storage pressures. Unlike scanners or servers, which generally involve one-time equipment purchases, storage requires continuous investment and ongoing capacity expansion. Some institutions have even had to delete valuable previously scanned data to free up space.
However, this does not mean we must wait until the storage problem is completely resolved before taking action. I have consistently advocated a “start small and move quickly” approach: begin with small specimens or specimens associated with specific diseases, whose storage requirements are more manageable, and gradually expand digitalization. This is a more practical and feasible pathway.
Overall, hardware-related challenges can be overcome. The key lies in changing the mindset of pathology professionals.
04
Oncology Frontier: As digital and intelligent breast pathology advances rapidly, what support is needed to help clinicians make better use of these tools and develop more precise treatment strategies for patients with breast cancer?
Professor Hong Bu: The development of pathology must always be grounded in the needs of clinical treatment, with the ultimate goals of supporting clinical decision-making and improving patient outcomes. Guided by this principle, digital pathology initiatives must be accompanied by a new professional mindset. The field should not remain confined to traditional morphological assessment and histological classification; instead, it should place greater emphasis on patients’ treatment needs and the evaluation of therapeutic outcomes.
Some of these emerging needs involve information that extends beyond what conventional microscopic pathological examination can provide. For example, digital and intelligent pathology may enable the deeper exploration of novel biomarkers and previously unrecognized morphological patterns. Such analyses could provide important evidence to guide treatment decisions, assess prognosis, and develop precision treatment strategies for individual patients.
Therefore, the development of digital and intelligent pathology is not merely a technological transformation. More importantly, it requires a fundamental shift in thinking, with far-reaching implications for both the advancement of the discipline and future academic research.
Professor Hong Bu
