Skip to main content

AI Model Spots 15 Missed Liver Tumors in Two Months

In a breakthrough that could change how liver cancer is diagnosed, Alibaba's DAMO Academy, in collaboration with Shengjing Hospital and China Medical University, has developed an AI model that can spot tiny liver tumors often missed by human eyes. The model, named DAMO LiON, was recently featured in the prestigious journal Nature Medicine.

During a two-month real-world trial, DAMO LiON assisted in reading CT scans for over 10,000 patients. It identified 15 malignant tumors that had previously gone unnoticed. Most of these lesions were about a centimeter in diameter—small enough to be easily overlooked. Thanks to the AI's detection, these patients received timely surgery or medication, potentially saving their lives.

The impact goes beyond just finding more tumors. The study showed that with AI assistance, doctors reduced their reading time by 27%, and their sensitivity to malignant tumors increased by 11.5%. Even more impressive, junior doctors were able to achieve the diagnostic level of senior doctors when using the AI tool. This could be a game-changer in regions where experienced radiologists are scarce.

So how does DAMO LiON work its magic? The model uses an improved network architecture that pays close attention to the relationship between the lesion and the entire liver, while also preserving local texture and boundary information. It can integrate different phases of enhanced CT images, capturing changes in small lesions at various scanning stages. This allows it to handle complex cases like fatty liver, liver cirrhosis, and post-operative livers with greater accuracy.

What makes this development particularly exciting is its potential to reduce missed diagnoses—a persistent challenge in cancer care. Liver cancer is one of the deadliest cancers worldwide, and early detection is crucial for survival. By catching these tiny tumors earlier, the AI model could significantly improve patient outcomes.

Of course, this is just the beginning. The team behind DAMO LiON plans to continue refining the model and expanding its use in clinical settings. As AI continues to evolve, we can expect more tools like this to assist doctors in making faster, more accurate diagnoses.

For now, the success of DAMO LiON serves as a powerful reminder of how technology can augment human expertise, making healthcare smarter and more accessible.

Key Points

  • DAMO LiON is an AI model developed by Alibaba's DAMO Academy and partners, published in Nature Medicine.
  • In a two-month trial, it identified 15 missed malignant liver tumors in over 10,000 patients.
  • AI assistance reduced reading time by 27% and increased sensitivity by 11.5%.
  • Junior doctors achieved senior-level diagnostic accuracy with AI help.
  • The model excels at detecting small lesions, even in complex cases like fatty liver or cirrhosis.