Researchers develop AI that can diagnose cancer

Researchers at Harvard Medical School have developed an AI called CHIEF that can diagnose cancer. An accuracy of almost 94 percent should enable timely treatment and improve the chances of survival.

Researchers at Harvard Medical School have developed a new AI called CHIEF. It is said to be able to diagnose and treat various types of cancer with high accuracy and predict survival rates. The artificial intelligence is similar to ChatGPT and analyzes digital tissue sections of tumors.

Based on cell characteristics, the system recognizes molecular profiles and survival indicators. With its ability to handle multiple tasks simultaneously, CHIEF could fundamentally change cancer diagnosis. The AI ​​can apparently not only recognize cancer cells, but also make predictions about treatment success and patient survival. This versatility could make the system a valuable addition to medical diagnostics.

AI diagnoses cancer – with 94 percent accuracy

The researchers trained the AI ​​with over 15 million images. They then tested the system on 19 different types of cancer. It outperformed other leading systems by up to 36 percent in areas such as detecting cancer cells, predicting tumor origins and identifying genetic mutations. These properties in turn influence treatment in many cases.

The AI ​​delivered similar results regardless of the method used to collect the tissue. In tests with over 19,400 images from 24 hospitals worldwide, it achieved an accuracy of almost 94 percent in diagnosing cancer. The hit rate was even higher when predicting genetic mutations that are crucial for treatment.

Innovation in tumor prognosis and treatment

CHIEF can apparently analyze not only the structure of tumor cells, but also the surrounding tissue. For example, the AI ​​identified features in the microenvironment of a tumor that indicate a better or worse response to standard treatments such as chemotherapy, immunotherapy and radiation.

What seems remarkable is that the AI ​​discovered features that were not previously associated with patient survival. The researchers plan to further improve CHIEF by training the model on rarer diseases and non-cancerous tissues.

They also want to feed the AI ​​with more molecular data to better detect more aggressive cancers and expand treatment options. If the system is further optimized and used worldwide, it could significantly improve cancer detection and treatment and offer personalized therapy options.

Also interesting: