A team of Chinese scientists has created an artificial intelligence tool that can predict the origins of tumors that are difficult to pinpoint accurately, making it competitive and possibly superior to human disease experts.
Doctors face a significant challenge in treating cancers of unknown origin due to the difficulty in diagnosing them. This type of cancer represents 3 to 5 percent of all cancer cases in humans and tends to be malignant, with an average survival rate of only 20 percent for patients suffering from it lasting just ten months.
Researchers, in collaboration with their colleagues from Tianjin Medical University and Chengdu University Hospital, have developed a new method based on deep learning to determine the origin of tumors. This method was trained using images of cancer cells from 57,220 cases in four Chinese hospitals.
The “TORCH” (TORCH) tool detects malignant tumors in fluid accumulated in the chest and abdomen and can predict their origin, as indicated by a study published this week in the journal “Nature Medicine.”
The tool showed an accuracy of 82.6 percent in predicting the origins of tumors, significantly enhancing the diagnostic accuracy compared to four human disease specialists, according to the study.
The overall survival period was improved to 27 months thanks to the treatment protocol according to “TORCH” (TORCH), compared to 17 months for those who received conflicting treatment.
Researchers noted that the research highlights the potential of artificial intelligence systems as a valuable tool in clinical practice, despite the need for further verification in randomized trials.
