Building an AI-Ready Digital Pathology Workflow: The Zhongshan Hospital Model with KFBIO and Huawei

Introduction: Moving Toward AI-Ready Digital Pathology

Digital pathology is entering a new stage driven by artificial intelligence (AI), automation and data intelligence. However, successful pathology AI requires more than advanced algorithms.

High-quality digital slides, standardized pathology data and reliable computing infrastructure are equally important.

Zhongshan Hospital, Fudan University (Shanghai), is developing a new model for intelligent pathology. The hospital combines high-throughput whole slide imaging, unified pathology data management, AI-assisted analysis and secure local computing into one clinical workflow.

This collaboration brings together three key strengths:

• Zhongshan Hospital, Fudan University (Shanghai) provides extensive clinical experience, pathology expertise and real-world validation.
• KFBIO (Konfoong Bioinformation Tech Co., Ltd)  provides digital pathology solutions, including the KF-PRX-XL high-throughput whole slide scanner, image management systems and pathology AI applications.
• Huawei provides the computing foundation for local AI deployment through the Huawei Ascend ecosystem.

Together, these capabilities create an integrated AI-ready pathology workflow. The goal is not only digital slide scanning. Instead, it connects slide production, image quality control, data management, AI model development and clinical applications.

The project represents a new approach to digital pathology transformation. It focuses on building a complete ecosystem where data, technology and clinical practice work together.

Why AI Pathology Starts with High-Quality Whole Slide Images

AI models depend on reliable input data. Poor image quality, missing tissue information or inconsistent staining can affect algorithm performance.

Therefore, laboratories must first establish a standardized whole slide imaging (WSI) workflow. This process converts glass slides into high-quality digital images that AI systems can analyze.

At Zhongshan Hospital, Fudan University, digital pathology is being developed as an end-to-end workflow. Scanned images are connected with pathology cases and managed through a unified platform.

As a result, digital slides can support multiple applications, including:

• Primary diagnosis
• Quality control
• Medical education
• Scientific research
• AI-assisted pathology analysis

The hospital has also adopted a practical approach to digital pathology implementation. It started digital primary diagnosis with gastrointestinal biopsy specimens and gradually moved the workflow from microscope-based review to screen-based review.

According to the pathology team, physicians adapted to the digital workflow within approximately one month. Traditional glass slides remain available for additional verification when required.

This transition highlights an important point: digital pathology becomes truly valuable when it becomes part of daily clinical practice.

An AI-ready laboratory is not built by simply storing digital images. Instead, it requires continuous integration between imaging, clinical workflows and AI applications.

The KF-PRX-XL: High-Throughput Imaging Engine for Large Pathology Laboratories

Large academic hospitals handle thousands of pathology slides every day. Therefore, they need more than excellent image quality.

They also require:

• High scanning capacity
• Stable operation
• Efficient slide management
• Consistent image output

To meet these needs, KFBIO developed the KF-PRX-XL high-throughput digital pathology scanner.

The KF-PRX-XL is designed for large-scale pathology environments. It supports high-volume scanning while maintaining image quality and workflow efficiency.

Key capabilities include:

• Up to 1200-slide loading capacity
• Brightfield scanning throughput of up to 200 slides per hour
• Continuous loading for 24/7 operation
• Slide pre-loading and automatic slide-order sorting
• Brightfield and fluorescence imaging
• 20x and 40x routine brightfield scanning
• Additional objective configurations for different applications
• Barcode recognition, slide sorting and intelligent archiving
• Support for KFB, SVS, DICOM and TIFF formats

Within the Zhongshan Hospital digital pathology architecture, the KF-PRX-XL serves as the high-throughput image acquisition engine.

It supports routine diagnosis and large-scale digitization. At the same time, it can work together with other scanners for urgent cases, specialized samples, fluorescence imaging and research workflows.

The value of high-throughput scanning goes beyond speed.

First, it creates a stable supply of standardized whole slide images for AI applications.

Second, continuous loading reduces manual intervention.

Finally, barcode tracking and case linkage improve traceability from physical slides to digital images and AI results.

From Scanning to an AI-Ready Pathology Data Platform

Whole slide imaging is only the first step in digital pathology transformation.

To support reliable AI applications, hospitals need more than digital images. They need a complete data workflow that connects images, clinical information and AI analysis.

KFBIO’s digital pathology platform connects scanned images with pathology case information. It supports centralized review, quality management, collaboration and research.

In an AI-ready pathology workflow, every case should meet several key requirements:

• A complete and high-quality whole slide image
• Accurate connection between patients, specimens and slides
• Structured clinical information required by AI models
• Proper authorization for clinical or research use
• Compatible specimen types and diagnostic applications
• Traceable AI results that can be reviewed by pathologists

These processes transform a large image archive into a valuable clinical data resource.

More importantly, they create a continuous improvement cycle for AI development.

For example, difficult cases can be identified, reviewed by experienced pathologists and added to curated datasets. These datasets can then support AI model validation and future optimization.

This approach is especially important in pathology.

Real clinical samples often contain complex conditions, including:

• Tissue folds
• Air bubbles
• Uneven staining
• Tissue loss
• Rare morphological patterns

AI systems must learn from these real-world challenges.

The pathology team at Zhongshan Hospital, Fudan University, has highlighted the importance of training AI models with practical clinical scenarios rather than only ideal images.

As a result, AI applications can become more reliable and closer to daily diagnostic needs.

What Huawei Ascend Contributes to the Pathology AI Infrastructure

Digital pathology generates massive amounts of data.

A single whole slide image can contain billions of pixels. Meanwhile, large pathology laboratories may generate thousands of images every day.

Therefore, AI pathology requires powerful and scalable computing infrastructure.

In this collaboration, the Huawei Ascend AI ecosystem provides computing support for pathology AI model development and deployment.

It helps address three major requirements:

1. Secure Data Management

Pathology images often contain sensitive medical information.

With local AI deployment, hospitals can keep valuable pathology data within their own controlled environment.

This approach supports data security and regulatory compliance.

2. Scalable AI Inference

Pathology AI applications require efficient image processing.

A scalable computing environment allows hospitals to analyze large numbers of digital slides without relying on individual workstations.

3. Continuous AI Optimization

AI models need continuous improvement.

With structured clinical data and expert feedback, pathology teams can validate models and improve performance over time.

For hospitals exploring locally deployed medical AI, the Zhongshan model provides an important reference.

The key value is not only the computing hardware.

Instead, it is the integration of:

• Whole slide imaging
• Digital pathology software
• AI algorithms
• Clinical validation
• Secure computing infrastructure

Together, these components create a complete AI-ready pathology ecosystem.

A Joint Laboratory for the Future of Intelligent Pathology

To further advance intelligent pathology innovation, Zhongshan Hospital, Fudan University, KFBIO and Huawei established the Joint Laboratory for Digital and Intelligent Pathology and Medical-Engineering Innovation.

The laboratory focuses on several key areas:

• Pathology foundation model development
• Intelligent automation for pathology laboratories
• Continuous pathology knowledge base development
• Clinical validation of AI applications

This collaboration demonstrates that clinical AI requires cooperation across different fields.

Hospitals provide:

• Clinical expertise
• Real-world cases
• Diagnostic requirements
• Validation standards

Digital pathology companies contribute:

• Whole slide imaging technology
• Workflow solutions
• Image management platforms
• Pathology AI applications

Computing partners provide:

• AI infrastructure
• Scalable computing resources
• Deployment capabilities

Together, these capabilities accelerate the development of practical AI pathology solutions.

AI Applications Designed for Clinical Pathology Work

The goal of pathology AI is not to replace pathologists.

Instead, AI is designed to support clinical decision-making, reduce repetitive tasks and improve workflow efficiency.

The Zhongshan Hospital project focuses on practical applications that can support daily pathology work.

Gastric Biopsy AI

Gastric biopsy diagnosis often requires careful review of large numbers of tissue regions.

The gastric biopsy AI application analyzes whole slide images and identifies suspicious areas for further review.

The model has been continuously improved through clinical data and expert feedback.

Its goal is to help pathologists:

• Find subtle abnormalities
• Prioritize important regions
• Improve review efficiency

However, final diagnosis remains the responsibility of qualified pathologists.

Cervical Cytology AI

Cervical cancer screening involves large numbers of samples.

AI can help laboratories improve screening efficiency by identifying slides that require closer examination.

The cervical cytology AI application helps:

• Prioritize abnormal cases
• Reduce repetitive screening work
• Allow experts to focus on complex cases

Through AI-assisted screening, pathology teams can improve resource allocation while maintaining diagnostic quality.

Quality Control and Continuous AI Model Improvement

AI output should not be treated as a final answer.

Instead, AI results should become part of a continuous review process.

In clinical pathology workflows, cases with uncertain results, unusual morphology or image-quality issues should enter an additional quality review process.

This workflow helps laboratories:

• Monitor AI performance
• Identify potential limitations
• Collect valuable clinical examples
• Improve future AI models

For pathology AI, continuous learning is essential.

Unlike laboratory test data, pathology images contain complex biological variations. Therefore, AI systems need ongoing validation with real clinical cases.

By combining AI results with expert review, hospitals can create a more reliable and practical AI-assisted pathology workflow.

What Overseas Hospitals Can Learn from the Zhongshan Model

The key lesson from the Zhongshan Hospital project is clear:

AI pathology should be built as a complete system, not purchased as separate tools.

A successful AI-ready pathology environment requires cooperation between hardware, software, data and clinical expertise.

A practical architecture includes four essential layers:

1. High-Throughput Whole Slide Imaging

Reliable image acquisition is the foundation of digital pathology.

High-performance scanners, such as the KFBIO KF-PRX-XL, provide the scanning capacity required for large pathology laboratories.

They help hospitals create consistent and high-quality digital slide resources.

2. Unified Digital Pathology Workflow

A digital pathology platform connects images, case information and clinical workflows.

It supports:

• Image management
• Remote consultation
• Quality control
• Research collaboration
• AI application integration

3. Clinically Validated Pathology AI Applications

AI models must solve real clinical problems.

They should support pathologists in tasks such as:

• Case prioritization
• Region identification
• Screening assistance
• Quality improvement

4. Scalable AI Computing Infrastructure

AI pathology requires secure and powerful computing resources.

Local AI deployment allows hospitals to maintain data control while supporting model development and clinical applications.

Hospitals can implement these capabilities step by step.

However, they should define operational goals from the beginning.

Important evaluation indicators may include:

• Digital primary diagnosis rate
• First-pass scan success rate
• Rescan rate
• Digital slide accessibility
• Viewer performance
• Reduction in physical slide retrieval
• Percentage of cases suitable for AI analysis

These measurements show whether digital pathology has truly become part of daily clinical practice.

They also help hospitals evaluate the long-term value of their digital transformation.

Building the Next AI-Ready Pathology Laboratory

The collaboration between Zhongshan Hospital, Fudan University, KFBIO and Huawei demonstrates a practical path toward intelligent pathology.

The journey starts with glass slides.

Then, it moves to standardized whole slide images.

Next, these images become valuable clinical data.

Finally, AI applications transform this data into practical diagnostic support.

KFBIO provides an integrated digital pathology solution that combines:

• High-throughput whole slide scanning
• Digital pathology workflow management
• Image storage and sharing
• Remote collaboration
• AI-assisted pathology applications

The KF-PRX-XL high-throughput scanner provides the imaging capacity required by large laboratories.

Meanwhile, Huawei Ascend provides scalable computing support for local AI deployment and model development.

Together, these technologies create a foundation for future pathology innovation.

For hospitals planning:

• Digital primary diagnosis
• Regional pathology networks
• AI pathology research
• Intelligent laboratory transformation

the Zhongshan model provides a valuable reference.

The future of pathology will not be defined by a single technology.

Instead, it will be built through the integration of imaging, data, AI and clinical expertise.

Frequently Asked Questions About AI-Ready Digital Pathology

What is the collaboration between Zhongshan Hospital, Fudan University, KFBIO and Huawei?

Zhongshan Hospital, Fudan University, KFBIO and Huawei established the Joint Laboratory for Digital and Intelligent Pathology and Medical-Engineering Innovation.

The collaboration combines clinical pathology expertise, digital pathology technology and AI computing infrastructure.

Its goal is to support pathology foundation models, laboratory automation and clinical AI applications.

What role does the KF-PRX-L play in AI pathology?

The KF-PRX-L is a high-throughput whole slide imaging scanner designed for large pathology laboratories.

It supports up to 640 slides, continuous loading and high-speed brightfield scanning.

By creating standardized digital slides, it provides reliable image data for pathology AI applications.

How does Huawei Ascend support digital pathology?

Huawei Ascend provides AI computing resources for pathology model development and deployment.

Within an integrated digital pathology workflow, it supports AI analysis, local deployment and continuous model optimization.

Does pathology AI replace pathologists?

No.

Pathology AI is designed to assist pathologists rather than replace them.

It can help identify areas of interest, prioritize cases and improve workflow efficiency.

Qualified pathologists remain responsible for diagnosis and clinical decisions.

What makes a pathology case suitable for AI analysis?

An AI-ready pathology case requires:

• A complete and high-quality whole slide image
• Accurate patient and specimen information
• Structured clinical data
• Appropriate usage authorization
• A supported specimen type and clinical indication

These elements help AI systems produce more reliable results.

Can this digital pathology architecture be used outside China?

Yes, the architecture can be adapted for international hospitals.

However, each region must consider:

• Local regulations
• Data protection requirements
• Product registration
• Clinical validation standards

Hospitals should select scanner, software and computing configurations based on their own workflow requirements.