Yitu Healthcare publishes AI diagnostic accuracy breakthrough
Contact Our Team
For more information about how Halldale can add value to your marketing and promotional campaigns or to discuss event exhibitor and sponsorship opportunities, contact our team to find out more
The Americas -
holly.foster@halldale.com
Rest of World -
jeremy@halldale.com
Yitu Healthcare has successfully used natural language processing (NLP) to achieve high-accuracy rates on par with doctors when reading electronic health records and generating patient diagnoses.
The result is published on Nature Medicine under “Evaluationand Accurate Diagnoses of Pediatric Diseases using Artificial Intelligence”.It’s the first time for a top medical journal to publish research findingsrelated to employing NLP technology in making clinical diagnoses, the companysays.
The system has achieved high accuracy of many dangerousconditions including acute asthma exacerbations (97 percent), bacterialmeningitis (93 percent), varicella (93 percent), influenza (94 percent) androseola (93percent). Yitu Healthcare joined hands with Guangzhou Women andChildren’s Medical Center and other research institutions to make thisbreakthrough.
Using Yitu’s deep learning based NLP system to extractclinically relevant information and establish a diagnostic system, the teamproposed a model for electronic health record data that integrates priormedical knowledge and data-driven modeling. The model has been applied in alarge pediatric population and shows a strong performance in accuracy acrossmultiple organ systems.
“It proves AI technology is able to assist doctors to dealwith large amount of data and diagnose, as well as provide support in uncertainand complex medical cases,” said Ni Hao, president of Yitu Healthcare.“Pediatric diseases can be tricky for doctors. An AI assistant will profoundlyimprove the diagnosis process and increase efficiency.”
The research has collected 101.6 million data points fromover 1.36 million outpatient visits from January 2016 to July 2017 in GuangzhouWomen and Children’s Medical Center to train and validate the model. Theprimary diagnoses included 55 diagnosis codes encompassing common diseases inpediatrics and representing a wide range of pathologies. The diagnostic systemachieved a robust performance for two categories of pediatric disease: commonconditions and dangerous potentially life-threatening conditions.
The system has great potentials for clinical use, forexample triage procedures. It generates predicted diagnosis with inputs ofbasic medical history, vital signs and physical exams from patients. Thesepredictions help doctors make better use of precious time. The system can alsoassist physicians to diagnose complex or rare conditions to avoid misjudgmentsor biases. Most importantly, the system will provide high-quality healthcareservice as well as ease the tension of severe shortage of experiencedpediatrics in China, benefiting both patients and doctors.
