Ruijin Hospital:
Transforming Pathology and Saving Lives with AI

Every year, nearly five million people in China receive a cancer diagnosis. For patients, the wait for a pathology report often feels unbearable, so pathologists race to offer accurate diagnoses, knowing all too well that time can be—quite literally—a matter of life and death.

Pathologists: "The Doctors' Doctors"

Down quietened corridors, the clock strikes 1 a.m. A pathologist at Shanghai's Ruijin Hospital rubs tired, reddened eyes. "People call us the doctors' doctors," the pathologist says, "but few realize that every pathology report is the result of hours of painstaking work at the microscope."

Seeing Through the "Camouflage of Cancer Cells"

The slide scanners in Ruijin Hospital's pathology department hum incessantly. Forty pathologists each handle over 300 slides every day, while consultations on complex cases can stretch across several, involving multiple expert eyes. This is the everyday reality for pathologists working here, in one of China's top-tier hospitals. "The camouflage of cancer cells is becoming increasingly sophisticated," one notes. "Some differential diagnoses feel like finding a single misplaced piece in a thousand-piece puzzle," And when manual inspection faces a flood of data, there are, of course, limits to human endurance.

Indeed, China has fewer than 20,000 pathologists in total—a number far from sufficient to meet the ever-growing demand for diagnoses. Even at primary hospitals, initial diagnoses are sometimes not comprehensive and fast enough, leaving patients waiting for answers. And this shortage of doctors, the heavy workloads, and the complexity of diagnoses have long been acknowledged as pathology's pain points. But the power of technology is beginning to transform all of this.

AI Microscopes: Making the Invisible Visible

In short, medicine is in the midst of a data revolution as pathology slide data volumes explode. Powered by Huawei's OceanStor Pacific scale-out storage, Ruijin Hospital and Huawei have built a healthcare data lake, where thousands of slides can be accessed in one second, historical specimens are preserved for up to 30 years, and unified compression algorithms save up to 45% of storage space. A pathologist demonstrates on a tablet: "In the past, retrieving a complex slide meant digging through the archives. Now, you just enter the number and instantly browse the slide where every trace of cancer is clearly visible."

Embedded within the Datacenter Virtualization Solution (DCS), Huawei's end-to-end toolchain ModelEngine powers Ruijin Hospital's RuiPath pathology model—turning AI into a super-assistant for doctors. Developed by Ruijin Hospital with Huawei's support, RuiPath is a clinical-grade, multi-modal pathology model that extracts vision features across typical cancer types, aligns cross-layer vision-language representations, and incorporates deep thinking mechanisms for long-sequence training.

Designed for end-to-end clinical pathology assistance, RuiPath currently covers 19 common cancers, representing 90% of all annual cancer cases in China, and supports hundreds of diagnostic tasks.

Through Huawei's ModelEngine, the entire process—from data preparation to model training, inference, and application orchestration—has become fully tool-based, standardized, and automated. What once required line-by-line annotation can now be batch-reviewed, significantly boosting efficiency while maintaining precision and professionalism. With ModelEngine's robust engineering, system interfaces are standardized and pathology report templates can be visually orchestrated. The rollout time for single-disease AI diagnostic applications has been cut from ten days to just two. And the role of doctors is evolving—from annotators to true reviewers. Empowered by AI, critical details once obscured are now revealed, measurable, and clear.

Beyond the Microscope

Fundamentally, at the heart of pathology lies the quest to resolve medical questions. Now, pathologists can see beyond the microscope, using digital and AI technologies, and that ultimately provides more answers, benefiting those that matter the most—the patients.