Researchers developed the TORCH deep-learning approach to differentiate tumor origins using cytological histology pictures from over 57,000 patients with hydrothorax and ascites. The AI model showed high accuracy in predicting cancer origin compared to pathologists. TORCH improved junior pathologists’ diagnostic abilities and guided treatment choices, leading to better overall survival rates in patients. The model performed well across various datasets, offering reliable predictions in clinical practice. However, issues like poor smear quality and image interpretation in specific cancer types need to be addressed for accurate results. The study demonstrates TORCH’s potential in guiding personalized treatments for cancers of unknown origin.
Source link
New AI tool ‘TORCH’ successfully identifies cancer origins in unknown primary cases
