Chinese scientists have developed an AI tool named TORCH that predicts tumor origins in cases of cancer of unknown primary (CUP) with high accuracy, rivaling or surpassing human pathologists. CUP presents challenges due to its elusive nature and poor prognosis. The deep-learning method, trained on cytological images from over 57,000 cases, can identify malignant tumors in fluids from chest and abdomen and predict their origins. TORCH achieved an 82.6% accuracy in predicting primary tumor origins, outperforming human pathologists. Treatment based on TORCH predictions led to better overall survival compared to discordant treatment. The study suggests TORCH could serve as a valuable tool in clinical practice, pending further validation in randomized trials.
Source – CGTN