Cancer of unknown primary (CUP) presents a challenge due to its poor prognosis and difficulty in determining the primary tumor site. Despite being heterogeneous, with metastases to various organs, diagnosing CUP is crucial for treatment. An artificial intelligence (AI) tool, TORCH, has been developed to assist pathologists in identifying the origin of metastatic cancer cells found in bodily fluids. Trained on thousands of images, TORCH accurately predicts the tumor source in 83% of cases, with a 99% chance of including the correct origin within its top three predictions. This technology can streamline diagnosis, potentially reducing the need for invasive tests and improving patient care.

Keywords: AI, metastatic cancer,  tumor source prediction, cancer diagnosis