Medical AI development is undergoing a fundamental shift from handcrafted, disease-specific models to factory-style mass production powered by foundational AI infrastructure. Hangzhou Diagens Biotechnology (02526.HK) has demonstrated this new paradigm by deploying 158 specialty-specific medical imaging models across nearly 100 hospitals in just six months, compressing what traditionally took years into a two to three month timeline.

What Is a Foundational Medical Imaging Model?

A foundational model is a large, general-purpose AI system trained on massive amounts of data that can be adapted for many specific tasks. In medical imaging, Diagens Tech developed iMedImage, a foundational model with 104 billion parameters trained on over 80 million medical images spanning 19 imaging modalities, from X-rays to MRI scans. Think of it like a master blueprint that can be customized for different medical specialties rather than building each specialty from scratch.

The key innovation lies in how this foundation enables rapid deployment. Once the foundational model exists, creating a specialty-specific model for, say, detecting lung cancer or analyzing cardiac imaging no longer requires years of data collection and training. Instead, teams use a process called fine-tuning, which adapts the general model to a specific medical task using smaller, focused datasets. This compression from years to months represents a productivity leap comparable to moving from artisanal manufacturing to industrial assembly lines.

How Does the Medical AI Factory Model Work?

• Data Standardization: Raw medical imaging data from different hospitals, equipment brands, and patient populations is cleaned and standardized through intelligent annotation and expert review. Diagens Tech has accumulated approximately 28.95 million annotated samples supported by over 3,000 specialized annotators as of mid-2026.
• Foundation Training: The standardized data trains a single, powerful foundational model once. This one-time investment supports unlimited downstream specialty-specific models without repeated high-cost training cycles.
• Rapid Fine-Tuning: New specialty models are created by adapting the foundation to specific diseases or organs using targeted datasets, compressing development cycles to two to three months compared to the years required for traditional approaches.
• Parallel Deployment: Multiple models can run simultaneously, and all benefit from improvements to the foundational model, creating network effects where the entire ecosystem becomes stronger with each update.

The financial impact is striking. In the first half of 2026, Diagens Tech's model service revenue reached 94.541 million Chinese yuan, representing a 101.1 percent year-on-year increase, while R&D spending grew only 67.4 percent. This divergence signals that the factory model is achieving economies of scale, where revenue growth outpaces investment growth.

Why Are Universities and Hospitals Embracing This Approach?

The shift toward foundational models is attracting academic partnerships. Hong Kong Polytechnic University and Diagens Tech jointly unveiled the PolyU-Diagens Joint Laboratory for Artificial General Intelligence and Medical Applications on September 2, 2026, signaling a move toward the next phase of medical AI development.

"AI for Science is reshaping the global medical AI landscape. AI presents challenges and opportunities comparable to the Apollo Program in helping humans decode life and health, and advance diagnosis, prevention and prediction of complex diseases," said Dr. Song Ning, Founder and Chairman of Diagens Tech.

Dr. Song Ning, Founder and Chairman, Diagens Tech

The partnership aims to push medical AI into what researchers call "Stage Three," the age of AI agents. Currently, even with foundational models, substantial manual work remains in data curation, parameter tuning, model training, and result analysis. AI agents would automate these workflows, further reducing the time and expertise required to develop new medical AI applications.

Dr. Li Yongqi, Project Lead of the Joint Lab, noted that there are over 5,000 medical imaging detection tasks globally awaiting solutions. The traditional small-model approach, where teams build dedicated models for specific diseases, cannot scale to meet this demand because each model requires years of development.

How Is This Reshaping Clinical Trials?

The factory model is also transforming how clinical trials capture medical data. Evinova, an AI-native clinical development platform, has partnered with Lothar Medical to integrate respiratory diagnostic devices with digital trial infrastructure. Lothar Medical's device combines three respiratory measurements, spirometry, oscillometry, and FeNO (fractional exhaled nitric oxide), into a single point-of-care platform.

This integration addresses a critical bottleneck in respiratory trials. Respiratory diseases are among the leading causes of death globally, yet diagnosis often occurs late in the patient journey. By combining multiple diagnostic capabilities in one device and connecting it to Evinova's unified trial platform, researchers can identify eligible participants more efficiently and capture high-quality endpoint data without data silos.

"Partnering with Evinova marks an important milestone for Lothar Medical and our commitment to advancing respiratory healthcare. By combining our innovative respiratory diagnostic capabilities with Evinova's digital clinical trial expertise, we can make clinical research more efficient, patient-focused and data-driven," said David Markus Thomas, Managing Director of Lothar Medical.

David Markus Thomas, Managing Director, Lothar Medical

The partnership is expected to deliver faster patient recruitment, reduced operational burden for sites and patients, and continuous, high-quality data capture for primary, secondary, and exploratory endpoints.

What Does This Mean for the Future of Medical AI?

The transition from handcrafted models to factory-style production represents a maturation of medical AI. Instead of each hospital or research institution building its own models from scratch, they can now access industrial-grade platforms that productize the entire workflow, from data governance through deployment and clinical feedback. This democratization of medical AI capabilities could accelerate the development of personalized medicine and improve diagnostic accuracy across diverse patient populations and disease areas.

The convergence of foundational models, academic partnerships, and integrated clinical trial platforms suggests that medical AI is entering a phase where scale, speed, and accessibility become competitive advantages. Organizations that master this factory model will likely lead the next wave of healthcare innovation.