Global challenges that the United States has yet to resolve have unexpectedly found answers from China.

At the First Affiliated Hospital of Zhejiang University’s radiology department, a team of over 300 staff handles nearly 9,000 patients daily, with close to 4,500 CT scans. The department head, with more than three decades of experience in medical imaging, offered a candid observation: "By afternoon, mental acuity begins to decline." This is not an isolated case of fatigue—it reflects a widespread reality across global radiology departments. With hundreds of millions of CT examinations conducted annually, the growth in physician numbers has failed to keep pace with the rising volume of scans.

This year, *Nature* published a dedicated article exploring "how AI can address radiology workforce shortages," its very title signaling urgency. On average, a single abdominal CT scan requires a radiologist to review 20 minutes’ worth of images and hundreds of slices—exactly the kind of workload repeated hundreds of millions of times worldwide each year.

The idea of using AI to supplement human labor is widely recognized, but progress has long been hindered by one persistent issue: current medical imaging AI systems are specialists. Each model is trained for a single condition—capable of detecting lung nodules, but not liver abnormalities. At best, they function as narrow-focused assistants, incapable of handling the full scope of a radiologist’s responsibility, which often involves assessing dozens of organs within a single report. What is needed is not another tool, but a generalist capable of taking on the entire workload of a radiology department.

China’s recent breakthrough precisely addresses this gap with a solution designed for versatility.

On September 18, DAMO RADAR—a joint project led by Alibaba’s DAMO Academy in collaboration with Zhejiang University’s First Affiliated Hospital and other institutions—was published in *Science*. The model demonstrates comprehensive capability across 18 abdominal organs and 146 diseases, achieving diagnostic performance on par with expert-level radiologists for the first time. In nearly 40,000 real-world clinical cases, it achieved an AUC score of 0.913. In a multi-institutional validation involving 26 radiologists from 14 hospitals, only three senior physicians slightly outperformed the AI. With AI assistance, radiologists saw a 10% improvement in diagnostic accuracy and reduced their analysis time by over 30%.

Effectively, this means each radiologist can produce higher-quality work more efficiently. For years, medical AI has promised to transform the industry. Now, that transformation is manifesting in every image read.

The urgency of the need becomes clear when examining real-world conditions. At Zhejiang University’s First Affiliated Hospital, the radiology department serves 9,000 patients per day, performing nearly 4,500 CT scans. Even after 30 years in the field, the department head admits, “by afternoon, mental sharpness fades.” This is exactly where AI must step in.

While *Nature* continues to debate whether such approaches are viable, China has already developed a general-purpose AI model that matches expert-level performance. Although still in the research and validation phase, word of the achievement has spread rapidly. Hospitals across the country are now awaiting its transition from academic publication into clinical practice. In regions suffering acute shortages of radiologists, waiting is no longer an option.

Original source: toutiao.com/article/1876650161554435/

Disclaimer: The views expressed in this article are those of the author alone.