AI vs doctors
A new artificial intelligence tool named PANDA is revolutionizing early detection of pancreatic cancer in China, potentially saving lives. The technology, developed by researchers affiliated with Chinese tech giant Alibaba, has already helped identify several cases of this deadly disease before symptoms appeared.
In a notable instance, Qiu Sijun, a 57-year-old retired bricklayer from eastern China, was unexpectedly contacted by Dr. Zhu Kelei, the head of the hospital’s pancreatic department, following a routine diabetes checkup. The call brought distressing news: Qiu had pancreatic cancer. Fortunately, the cancer was detected early, allowing Dr. Zhu to successfully remove the tumor.
The breakthrough came from the implementation of PANDA, short for “pancreatic cancer detection with artificial intelligence.” This tool is designed to analyze non-contrast CT scans, which are less invasive than traditional scans but typically provide less clarity. Traditionally, diagnosing pancreatic cancer has been challenging due to its late onset of symptoms and the risks associated with high-radiation imaging.
PANDA has been undergoing clinical trials at the Affiliated People’s Hospital of Ningbo University since November 2024. The tool has processed over 180,000 abdominal or chest CT scans, leading to the detection of approximately two dozen cases of pancreatic cancer, including 14 early-stage diagnoses, according to Dr. Zhu. Among these cases were 20 instances of intraductal adenocarcinoma, the most common and aggressive form of the disease.
Dr. Zhu emphasized the importance of this tool, stating, “I think you can 100% say AI saved their lives.” The ability to detect cancers early is crucial, as pancreatic cancer has a five-year survival rate of only around 10%, primarily due to late diagnosis.
In April 2024, PANDA received “breakthrough device” status from the U.S. Food and Drug Administration, expediting its review process for market entry. Despite its promise, researchers caution that additional data is needed to fully assess the tool’s effectiveness and minimize risks of false positives.
While PANDA has shown impressive results, experts outside the project have expressed skepticism about the reliability of non-contrast CTs compared to other imaging methods. Ling Zhang, a senior staff algorithm engineer at Alibaba’s Damo Academy and one of PANDA’s creators, acknowledged these concerns. To enhance accuracy, the team used data from over 2,000 known pancreatic patients, allowing the AI to learn how to detect potential cancers in less detailed images.
A study published in Nature Medicine in 2023 confirmed that the tool could accurately identify 93% of individuals with pancreatic lesions from non-contrast CT scans. At the Ningbo hospital, scans flagged by PANDA undergo review by doctors, who then follow up with patients when necessary. This method ensures that no additional costs are incurred for patients, as many undergo non-contrast CTs as part of their routine checkups.
Despite these advancements, challenges remain. The hospital struggles with staffing limitations for follow-ups, and outdated hardware often hampers the system’s performance. Additionally, the potential for false alarms raises concerns. Dr. Ajit Goenka, a radiologist at Mayo Clinic, highlighted the psychological burden false positives can impose on patients, stressing the importance of minimizing unnecessary callbacks.
Dr. Diane Simeone, a pancreatic surgeon at the University of California San Diego, noted that while PANDA could serve as a valuable tool, especially for less experienced doctors, the effectiveness of AI in detecting obvious tumors is still a matter of debate. Some detected tumors should ideally be identifiable by well-trained radiologists without AI assistance.
The technology is also being tested in rural areas, such as a clinic in Yunnan Province, where access to specialists is limited. In Ningbo, the apparent success of PANDA has generated new demands on hospital resources, with Dr. Zhu highlighting the need for increased staffing to handle the follow-up workload.
Public trust is another significant factor influencing the tool’s adoption. In a healthcare landscape marred by corruption, some patients may be reluctant to seek follow-ups due to concerns about financial motives. Nevertheless, Qiu Sijun was not deterred; he opted for surgery following Dr. Zhu’s recommendation. During a follow-up visit, Qiu expressed relief at his recovery and a newfound appreciation for the healthcare he received.
“I was very lucky,” he remarked, after learning about his early-stage cancer detection. The development of PANDA signifies a crucial step towards improving outcomes for patients facing one of the most challenging cancers, illustrating how technology can bridge the gap in early detection and diagnosis.
