NEC Corporation (NEC) and Tohoku University Hospital have conducted a demonstration experiment to automatically create medical documents using a Japanese large-scale language model (LLM) in generative AI, which has the potential to cut medical document creation time in half and improve operational efficiency. It was confirmed. NEC and Hashimoto Municipal Hospital (Hashimoto City, Wakayama Prefecture) are also conducting similar demonstration experiments starting in October 2023.
In the medical field, the shortage of manpower is becoming more serious, and the burden on doctors and nurses is increasing. On the other hand, with the implementation of a new system for ``physician work style reform,'' an upper limit will be placed on the overtime hours of working doctors. DX (digital transformation) of medical operations is expected to improve efficiency.
Tohoku University Hospital utilizes medical text analysis AI developed by NEC to organize information such as patient symptoms, test results, progress, and prescriptions in electronic medical records in chronological order. Using NEC's LLM, it was possible to automatically generate a summary of the treatment progress.
In the hospital's demonstration (October to November 2023), the time required to create summaries to be included in referral letters and discharge summaries was reduced by an average of 10% compared to creating new ones. The expression and accuracy of the sentences were also highly evaluated. As a result, the work of collecting information from vast amounts of electronic medical records has been significantly reduced, and the possibility of efficient document creation from the generated summary text has been confirmed.
In this demonstration at Hashimoto City Hospital, anonymized information from electronic medical records is safely and seamlessly linked to LLM on the cloud via NEC's electronic medical record system, and a summary text is generated while avoiding learning of personal information. . In the future, the company aims to generate summaries of long-term treatment progress and improve operability.
In the future, NEC will also support the creation of medical texts using LLM + voice recognition and demonstrate automatic creation of detailed symptom descriptions (supplementary explanations added to medical records) that place a heavy burden on doctors. The company also plans to implement a function that automatically generates a summary statement by simply clicking a button on the electronic medical record.
