Alibaba's AI Model cancerUniT Boosts Cancer Diagnosis Accuracy
Alibaba's Damo Academy has developed an AI model, cancerUniT, aiming to enhance cancer diagnostic accuracy and reduce errors. The model can detect and differentiate eight common cancers, including lung, colorectal, liver, and breast, along with specific tumor subtypes.
cancerUniT uses the Transformer neural network to learn unique traits and patterns of each tumor type and their relationships within organs. Built on Mask Transformer, a semantic segmentation technique, the model analyzes images of body tissues at a pixel level.
The model has shown promising results in a study involving 631 patients. It demonstrated higher sensitivity (93%) and specificity (82%) in detecting and segmenting tumors compared to traditional models. This means cancerUniT can identify more cases without false alarms, aiding radiologists in their diagnoses.
Damo Academy collaborated with renowned medical institutes, such as the Chinese Academy of Sciences, Peking University Cancer Hospital, and Beijing Union Medical College Hospital, to develop this multi-cancer image analysis model.
cancerUniT simplifies and improves the diagnostic process for radiologists, helping them identify recurring or spread cancers. With its high sensitivity and specificity, the model has the potential to revolutionize cancer diagnosis, potentially saving lives and reducing healthcare costs.
