Cancer cachexia is a serious problem that affects up to 80% of people with advanced cancer. It causes major weight loss, muscle weakness, and tiredness. Sadly, just eating better is not enough to fix it.
Even though cachexia is common, it’s hard to diagnose. It can lower the quality of life, make treatments less effective, and shorten survival. Often, it isn’t found until it’s very advanced.
Diagnosing cachexia is tricky because its symptoms can look like side effects from cancer treatment or other problems like malnutrition. Also, weight loss is not always tracked well, and people may feel uncomfortable talking about it. Plus, there aren’t many standard tests to detect it early.
Now, researchers at Moffitt Cancer Center have created a new AI tool that could make a big difference. This tool uses imaging scans and regular health data to spot cachexia earlier and more accurately. They shared their findings at a big cancer research meeting.
Graduate student Sabeen Ahmed, who worked on the study, said, “Cancer cachexia often goes unnoticed until it’s too late. Our AI tool uses different types of data to catch it earlier, helping slow down muscle loss and improve health.”
The AI model looks at CT scans to measure muscle mass. It combines this with lab results, medical records, and other health information to predict if a patient has or will get cachexia. In one study, it correctly identified cachexia in 77% of pancreatic cancer patients. When lab results and doctor’s notes were added, the accuracy rose to 85%.
The AI also showed good results when used on patients with pancreatic, colorectal, and ovarian cancers, improving diagnosis accuracy by 6.7%, 3.0%, and 1.5%, compared to using regular clinical data alone.
Ahmed explained that using many types of data is key to the tool’s success. “Each piece of data is like part of a puzzle,” she said. “Together, they help us see patterns that single tests can miss.”
The AI’s muscle measurements were very close (within 2.48%) to what expert doctors found manually. It also shows how confident it is in its results, making it more reliable for real-world use.
The study focused mainly on a few types of cancer and only used CT scans. But researchers believe adding PET or MRI scans and expanding it to more cancer types could make the tool even better.
Ahmed said, “Our work shows how AI can change cancer care by helping doctors create better, more personal treatment plans and improve patient health.”
Source: moffitt
