AI is Finding Deadly Tumors that Doctors Might Miss

A New York hospital in eastern China is quietly piloting an A.I. system called PANDA

 that is catching pancreatic cancers on routine CT scans long before human doctors might 

notice them, giving some patients a rare chance at curative surgery. The article by

 Vivian Wang uses this project to illustrate both the life‑saving potential and the 

ethical and practical challenges of weaving advanced A.I. into everyday medical 

care.

​

What PANDA Actually Does

• PANDA (“pancreatic cancer detection with artificial intelligence”) scans 

  non‑contrast abdominal and chest CT images that patients are already getting 

  for other reasons, and flags subtle abnormalities that could be early pancreatic 

  tumors.​

• In one hospital in Ningbo, the tool has reviewed more than 180,000 scans and 

  helped uncover about two dozen pancreatic cancers, including 14 caught at an

   early, more treatable stage.​

Why This Matters for Pancreatic Cancer

• Pancreatic cancer is among the deadliest cancers worldwide, with a five‑year 

  survival rate around 10 percent largely because it is usually discovered late, after symptoms appear and the disease has spread.​

• Standard high‑resolution imaging is too invasive and costly for mass screening,

   while cheaper non‑contrast CTs are harder for radiologists to read—precisely 

  the gap PANDA aims to fill by extracting more information from those 

  lower‑quality scans.​

How The System Was Trained

• Engineers linked to Alibaba’s research arm worked with radiologists to label tumor

   locations on high‑quality contrast CT scans from more than 2,000 patients, 

  hen mapped those labels onto matching non‑contrast scans to teach the model 

  what to look for in fuzzier images.​

• In a study published in Nature Medicine, PANDA reportedly identified about 

  93 percent of pancreatic lesions in a test set of over 20,000 non‑contrast CT

   scans, a performance level that surprised even its creators.​

Hopes, Doubts, and Risks

• Some cancer specialists quoted in the article say the tool can be especially 

  valuable in hospitals that lack highly trained pancreatic experts, effectively 

  acting as a second set of eyes on every scan.​

• Others warn that false positives can generate fear, unnecessary invasive tests,

   and higher costs, and note that some of the tumors flagged by the system 

  should have been “obvious” to an experienced radiologist even without A.I.​

Broader Context and Human Stories

• The U.S. Food and Drug Administration has granted PANDA “breakthrough device”

   status, fast‑tracking its regulatory review, while multiple clinical trials continue in

   China to see whether the real‑world benefits outweigh the downsides.​

• The article closes with patients like a farmer whose early‑stage pancreatic tumor 

  was removed after PANDA flagged his scan; he admits he does not understand

   how A.I. works, only that a hidden cancer was found in time and his life may 

  have been saved.​

AI Summary:

 A Chinese research team has built an A.I. system called PANDA to spot pancreatic

cancer early by scanning routine CT images that patients are already getting for 

other reasons. The tool analyzes non‑contrast scans, which are cheap but harder

for humans to read, and flags tiny abnormalities that might be early tumors. 

In a major hospital trial, PANDA reviewed more than 180,000 scans and helped 

find dozens of cancers, including many at a stage when surgery is still possible. 

Doctors see huge promise because pancreatic cancer is often detected too late 

and has one of the lowest survival rates worldwide. At the same time, experts warn

 about false alarms, extra tests, cost, and the risk of over‑relying on algorithms. 

The article ends with patients whose hidden cancers were caught in time, 

highlighting both hope and uncertainty as A.I. moves deeper into medicine.​