An examiner checks the results made by an artificial intelligence (AI) model for blood cell testing in a laboratory at Xinqiao Hospital under the Army Medical University in southwest China's Chongqing, Feb 14, 2023. (PHOTO / XINHUA)

SHANGHAI -- A Chinese team has developed an artificial intelligence (AI) platform that can help detect early-stage pancreatic cancer using X-ray pictures with high accuracy.

PANDA, a deep learning approach developed by Alibaba's Damo Academy, can magnify and identify subtle pathological features in plain CT images that are difficult to detect with the naked eye.

PANDA was initially trained by using a dataset of 3,208 patients from a single center. According to a study published this week in the journal Nature Medicine, the AI tool achieved a sensitivity of 94.9% and a specificity of 100% in an internal test cohort of 291 patients from the Shanghai Institution of Pancreatic Disease

Pancreatic cancer is a highly malignant tumor, but it has the potential to be curable if detected very early.

READ MORE: Pig organs transplanted into monkeys in attempt to sew up shortage of donors

PANDA was initially trained by using a dataset of 3,208 patients from a single center. According to a study published this week in the journal Nature Medicine, the AI tool achieved a sensitivity of 94.9 percent and a specificity of 100 percent in an internal test cohort of 291 patients from the Shanghai Institution of Pancreatic Disease.

READ MORE: AI virtual hosts change game for livestreaming

The model was then validated on external multicenter cohorts of 5,337 patients from China and the Czech Republic using abdominal CT scans. Its sensitivity and specificity were 93.3 percent and 98.8 percent, respectively, outperforming the performance of the average radiologist.

READ MORE: China's AI robotic chemist finds way to make oxygen on Mars

Sensitivity refers to a test's ability to designate an individual with disease as positive, while the specificity of a test is its ability to designate an individual who does not have a disease as negative.

The PANDA model has been used more than 500,000 times clinically, with only one false positive occurring every 1,000 times.

PANDA could potentially serve as a new tool for large-scale pancreatic cancer screening, said the researchers.