The liver cancer diagnostic AI model DAMO LiON, developed by Alibaba DAMO Academy in collaboration with institutions such as Shengjing Hospital, China Medical University, has recently been published in the top international medical journal Nature Medicine. The model can identify small liver lesions through enhanced CT images, especially those that are easy to miss, such as liver metastases.
In a two-month real-world prospective clinical trial, DAMO LiON assisted in reading images for more than 10,000 patients, identifying 15 malignant tumors that had been previously missed. Most of these lesions were approximately 1 cm in diameter and enabled patients to receive timely surgery or medication treatment.
Research shows that after using AI-assisted image reading, doctors' reading time was reduced by 27%, and their sensitivity to malignant tumors increased by 11.5%. At the same time, junior doctors were able to achieve the diagnostic level of senior doctors with AI assistance.
According to the introduction, DAMO LiON uses an improved network architecture, paying attention to the relationship between lesions and the entire liver, while retaining local texture and boundary information. It can also integrate different phases of enhanced CT images, capturing changes in small lesions at different scanning stages, thereby improving the ability to identify complex cases such as fatty liver, liver cirrhosis, and postoperative livers.
