AI Developed in China Capable of Reading Cancer Images
The model developed by Chinese researchers demonstrated over 95% accuracy in nearly 50 cases.
Researchers from the Air Force Medical University in China, Tsinghua University, and AI technology developer SenseTime have developed a versatile AI model capable of detailed reading of cancer pathological images. The study used approximately 300,000 digital pathology images compiled from local data in the country.
The research indicated that the AI model could analyze more than 20 organs. These analyses include pancreas classification, lesion definition, multiple cancer assessments, and biomarker evaluations.
Accuracy Rate of 95%
The versatile AI model, named PathOrchestra by the researchers, achieved an accuracy rate of over 95% in nearly 50 cancer cases. Commentators on the research noted that this study also reflects China’s rapid growth in the field of artificial intelligence.
The AI model trained itself using a method called cross-learning. Professor Wang Zhe from the Academy of Basic Medical Sciences at the Air Force Medical University stated that perceiving and interpreting pathological images is very challenging due to the differences among pathogens, making the examination of diseased structures one of the most difficult topics in image processing.
If further developed, this model is expected to significantly reduce the burden on pathologists.
