Cancer is a terrible and serious disease. However, one of its types can be worse than cancer – cancer of unknown origin, which occurs in 3-5% of cases.
The problem with this type is that it usually manifests itself in the late stages with metastases, and the unknown origin of the tumor complicates the application of any treatment because their different types require completely different methods. At the same time, currently known methods for diagnosing tumors show very low accuracy in determining their origin.
Artificial intelligence has become a new possible solution to the problem. Using machine learning techniques, the researchers trained the AI model on tens of thousands of cancer samples. As a result, it achieved an impressive diagnosis accuracy of 97% for known types of cancer, and 72% for cancers of unknown origin.
Moreover, in 40% of cases, AI corrected a misdiagnosis made by a person.
