Challenge

The Endoscopy AI Group within the medical AI R&D division at a leading medical research institution faced major challenges in analyzing endoscopic images for cancer research. As the team worked with massive volumes of endoscopic images, it needed to solve several critical issues:

• It had to accurately identify and analyze tumor regions across hundreds of thousands of endoscopic images.
• Previously, medical staff had to manually label tumors using bounding boxes, a process that consumed enormous time and manpower.
• The team needed a way to automatically distinguish normal regions, tumor regions, and endoscopic tool regions within each image for precise analysis.
• There was no function to automatically calculate the proportion of normal tissue in the full image, excluding tumor and tool regions.
• Because the project involved sensitive medical data, strict security controls were essential.

These challenges slowed the pace of research and increased the need for a more precise and scalable analysis environment.

Solution

To address these issues, the institution adopted Superb AI’s on-premise AI platform. Superb Platform provided the following core capabilities:

• Advanced auto-labeling: Superb AI’s platform introduced advanced auto-labeling technology capable of automatically labeling approximately 1 million endoscopic images. This dramatically improved the previous manual labeling workflow.

(Auto-labeling interface in Superb Platform)

• Precise segmentation model: Using a segmentation model based on SAM (Segment Anything Model), the platform accurately separated normal regions, tumor regions, and endoscopic tool regions within endoscopic images.
• Area calculation capability: The platform automatically calculated the proportion of normal tissue within the full endoscopic image, enabling more accurate measurement of tumor density and size.
• Integrated AI development environment: Data curation, model training, model diagnostics, and generative AI capabilities were integrated into a single platform, creating a more efficient research environment.
• Security-enhanced on-premise deployment: The platform was deployed on-premise, allowing the institution to process sensitive medical data securely within its own environment without exposing data externally.
• Professional support system: Superb AI provided continuous technical support and consulting, helping the research team maximize the value of the platform in its workflow.

Benefit

After adopting Superb AI’s platform, the institution achieved several notable outcomes:

• Improved data processing efficiency: By automatically labeling hundreds of thousands of endoscopic images, the team significantly reduced research time and was able to redirect human resources toward higher-value research activities. Previously, the institution had relied on an open-source labeling tool, Product C, which required staff to manually label each endoscopic image one by one. Tasks that had previously taken several minutes per image were replaced by Superb’s auto-labeling technology, reducing processing time to as little as one second per image and improving efficiency by several dozen times.
• Higher analysis accuracy: With a precise segmentation model, the team was able to distinguish tumor regions and normal regions more accurately, while the area calculation function made it possible to measure tumor density objectively. Under the previous manual workflow, improving analysis accuracy required an additional review process after labeling, which often added another one to two days. With Superb’s auto-labeling technology, validation became possible immediately, and region detection accuracy improved significantly beyond the manual labeling baseline.
• Stronger medical data security: Through the on-premise platform, the institution was able to maintain strict security over sensitive medical data while still taking full advantage of advanced AI capabilities.
• Greater research scalability: The institution, recognized among the world’s top 15 in oncology, is actively applying AI to cutting-edge cancer research. To analyze massive volumes of lesion images, it selected Superb Platform, which can also be extended beyond endoscopic images to other large-scale image analysis use cases such as X-ray imaging. With robust capabilities across data curation, model training, model diagnostics, and generative AI, the platform enables efficient and detailed analysis in a single environment, making it applicable across multiple areas of cancer research.
• Expanded clinical applicability: The AI model developed through this project can be used not only for research publication, but also for follow-up treatment of patients, significantly increasing the practical clinical applicability of the research outcomes.

This case is a strong example of how advanced AI can drive meaningful innovation in medical research. Superb AI’s platform goes beyond providing a technical tool. It improves the efficiency and precision of medical research and ultimately contributes to better patient care outcomes.