Google’s MedGemma 1.5 can interpret multimodal medical images, including CT and MRI scans and pathology slides. OpenAI’s health features can answer questions by combining medical records with sleep and activity data. Yet as the technical capabilities of medical AI advance, a fundamental tension remains: medical AI has no fixed set of “standardized” answers. Clinicians can read the same scan differently. New disease entities and diagnostic modalities continue to emerge, meaning that even today’s most advanced models may need to be rebuilt within a few years. What lies ahead for medical AI?

On Sept. 25, during the fifth Global Digital Trade Expo, Dr. Ning Song, founder and chairman of the board of Diagens Tech, had an in-depth conversation with Kevin Kelly, the founding editor-in-chief of Wired magazine. Their conversation brought a shift into focus: as the line between AI and clinicians is redrawn, the competitive dynamics of medical AI are quietly changing.

What Is the Real Competitive Moat in Medical AI?

As AI takes on more medical information processing and diagnostic tasks, the division of work between AI and clinicians is being redefined.

Kelly believes AI outperforms humans in providing medical information and performing diagnostic tasks, while clinicians bring empathy and the capacity for breakthrough innovation—capabilities AI cannot replicate. Amid rapid technological progress, clinicians need to keep learning throughout their careers, while AI systems must continue to evolve. Customizable medical imaging AI developed for individual hospitals and clinicians allows clinicians to play a greater role in developing the technology, feeding their professional expertise into ongoing model training.

That view aligns closely with Diagens Tech’s approach in practice. The company believes medical AI will thrive through collaboration between humans and AI. In the future, radiologists will be able to solve their own specialized clinical challenges and train tailored AI copilots based on a shared foundation model.

This “clinician-as-developer” model points to a shift in the medical AI business model, from selling software to providing capabilities. Software will become obsolete, but capabilities can be reused and monetized on an ongoing basis. This may be a key way for medical AI companies to build a lasting competitive moat.

A Full-Stack Product Suite: From Data Accumulation to Clinical Deployment

At the fifth Global Digital Trade Expo, Diagens Tech showed how medical imaging data can drive AI-enabled productivity. Its product suite has four layers: iMedImage® is the foundation model for medical imaging; iMedStudio™ lets specialists participate in data annotation, correction and quality control; iMedMaaS® supports the training and deployment of models for specific specialties; and iMedLoop™ links data collaboration, evaluation, release and feedback.

Specifically, iMedStudio™ handles medical image processing, AI-assisted annotation, expert review and quality control. The platform now has more than 3,000 professional annotators and a high-quality medical imaging dataset of 28.95 million records.

iMedLoop™ comprises four core modules: iMedStudio™, a workspace for multimodal intelligent annotation and data production; iMedImage®, a large-scale foundation model for medical imaging; iMedMaaS®, a model-training and deployment platform; and DoctorBench®, a unified evaluation system. Together, these modules facilitate secure data collaboration across institutions.

Building on this infrastructure, Diagens Tech showcased more than 20 representative use cases in obstetrics and gynecology, growth and development, cardiovascular and cerebrovascular conditions, emergency and critical care, and oncology. They span imaging modalities including CT, MRI, X-ray, ultrasound, CTA and endoscopy. All draw on the underlying capabilities of iMedImage®, the universal foundation model for medical imaging.

Diagens Tech describes its end-to-end pipeline—from accumulating medical imaging data and generating high-quality data to developing and training models and deploying them clinically—as an accelerator for medical imaging AI research and production. Rather than rely on any single model’s superiority, it uses closed-loop workflows to build a production system that can be updated continuously, directly addressing medical AI’s fundamental challenge: clinical scenarios are neither standardized nor static.

To date, Diagens Tech has made substantial progress in regulatory compliance and commercialization. AI AutoVision®, its karyotype-assisted diagnostic software built on iMedImage®, received Class III medical device registration from China’s National Medical Products Administration (NMPA) this year. The iMedImage program covers 43 human organs or anatomical sites and 61 disease areas. Diagens Tech has established partnerships with 99 hospitals, including 65 Grade A tertiary hospitals.

Looking ahead, Kelly says a company’s real long-term core asset may be neither its software nor its algorithms, but a repeatable process for creating new software, new algorithms and, above all, new medical imaging AI systems.

This marks a fundamental shift in how medical AI companies compete. In the past, the industry competed to build the most advanced models. In the future, competition will center on the ability to build models tailored to new use cases continuously, efficiently and in a compliant manner. Diagens Tech’s universal foundation models and suite of tools fit this transition perfectly. With reusable foundation layers, accumulated data and closed-loop workflows, every deployed project fuels further expansion, creating a self-reinforcing platform flywheel that grows stronger with use and drives exponential business growth. If this roadmap succeeds, Diagens Tech is poised to become a platform leader by 2030, with annual revenue in the RMB 10 billion range, net profit of RMB 3 billion and market capitalization exceeding RMB 200 billion. Its ability to keep evolving represents the sector’s scarcest valuation anchor today.