Chinese startup promises 97% accuracy in AI pediatric diagnosis
Innovation Enterprise | February 14, 2019
YITU Healthcare, a Chinese AI startup, has claimed to have developed a system that uses natural language processing (NLP) to achieve an accuracy of up to 97% when diagnosing pediatric diseases. This puts the AI's capabilities on par with a trained doctor's when reading electronic health records and forming a diagnosis. The company trained the system using 101.6 million data points from more than 1.36 million patient visits over an 18-month period. The NLP-enabled system is supported by a data model which integrates clinically relevant information and prior medical knowledge. It is able to diagnose a number of common pediatric diseases, recording an 97% rate diagnosing acute asthma exacerbations, 93% when diagnosing bacterial meningitis in addition to varicella (93%), influenza (94%) and roseola (93%). Ni Hao, President of YITU Healthcare, explained that using AI can help doctors use large amounts of health data as well as helping them diagnose complex medical cases.