Skip to main content

AI Model Spots Tiny Liver Tumors Doctors Missed

In the fight against liver cancer, early detection is everything. But the liver is a tricky organ—small tumors can hide among the shadows of cirrhosis, fatty liver, and complex anatomy. Even experienced radiologists can miss them. That's where Alibaba's DAMO Academy steps in with a new AI model called DAMO LiON.

Developed in collaboration with Shengjing Hospital and other institutions, DAMO LiON is designed to spot small liver lesions on enhanced CT images. In a real-world prospective trial lasting just two months, the model flagged 15 malignant tumors that had been previously overlooked. Most of these were around 1 centimeter in size—tiny enough to slip past even the most careful human eye.

The findings, published in Nature Medicine, highlight the potential of AI as a safety net in medical imaging. Think of DAMO LiON as an "AI safety officer"—it doesn't replace doctors but works alongside them, double-checking scans and catching what might be missed.

A Second Pair of Eyes

Liver cancer and liver metastases are notoriously difficult to detect early. Small lesions are easily obscured by underlying conditions like cirrhosis or fatty liver, and doctors' attention is often drawn to the primary tumor, leaving secondary growths unnoticed. DAMO LiON aims to fill those gaps.

The model uses an improved network architecture that balances the relationship between lesions and the entire liver while preserving local texture boundaries. It can also integrate images from different phases of the CT scan, capturing fleeting details that might otherwise be lost.

In tests, DAMO LiON showed higher accuracy in identifying malignant tumors than radiologists alone. When used together with human readers, it reduced reading time by 27% and increased sensitivity by 11.5%. Junior doctors, with the AI's help, could perform at the level of senior experts.

Real-World Impact

But how does it work in practice? In the trial, if the AI's initial diagnosis conflicted with the first assessment, senior doctors reviewed the case. If needed, the case was escalated for multidisciplinary discussion. Over two months, the AI analyzed more than 10,000 patients and identified 15 liver metastases that had been overlooked.

One striking case involved a 65-year-old bladder cancer patient. Her tumor markers were normal, and the initial diagnosis showed only calcification. But the AI flagged a liver metastasis, leading to a change in her chemotherapy treatment. That's a life-altering catch.

A Decade of Dedication

Since 2017, DAMO Academy has been focusing on medical AI, expanding its capabilities from screening to diagnosis. The goal is simple: to enable more patients to receive timely and accurate treatment. With tools like DAMO LiON, that goal is becoming a reality.

As AI continues to evolve, it's not about replacing doctors—it's about giving them superpowers. And for patients, that could mean the difference between a missed tumor and a timely intervention.

Key Points

  • DAMO LiON is an AI model developed by Alibaba's DAMO Academy and Shengjing Hospital for liver cancer diagnosis.
  • It can detect small liver lesions on enhanced CT images, including tumors as small as 1 cm.
  • In a two-month trial, it identified 15 overlooked malignant tumors in over 10,000 patients.
  • The model improves sensitivity by 11.5% and reduces reading time by 27% when used with radiologists.
  • The research was published in Nature Medicine.