BRIN Introduces LISA V2, AI-Based Medical Chatbot For Hemodialysis Patient Education
JAKARTA - The National Research and Innovation Agency (BRIN) introduced LISA v2 at the 12th IC3INA international conference. This chatbot is specially designed for health education needs for hemodialysis patients.
This innovation was presented by Research Center for Data and Information Science Research (PR SDI) BRIN Siska Pebiana in a conference held in a hybrid manner. The conference highlighted various researches on Artificial Intelligence (AI) and big data.
"LISA v2 is not just a chatbot, but an educational partner for patients to understand its condition in easy language and valid data," said Siska, quoted on the official BRIN website on Tuesday, October 21.
LISA v2 is developed using the latest Large Language Model (LLM) and Retrieval Augmented Generation (RAG) technology. This technology is combined to produce structured and scientific references supported answers.
Unlike medical chatbots in general, LISA v2 can answer complex questions with credible references. The trusted sources used as references include Halodoc, Hello Sehat, Scopus, and Google Scholar.
The latest version of this chatbot also involves medical personnel in the validation process to ensure the accuracy of all information provided. This is done to maintain the quality and safety of health information submitted to patients.
SEE ALSO:
The LISA v2 system consists of three main components, namely verified medical data collection, RAG-based vector database, and LLM chatbot machines. This system also features different interfaces for the public, medical personnel, and administrators.
In the preliminary test using 70 data from doctors, the Sahabat AI model at LISA v2 recorded the highest accuracy compared to other popular AI models. Models made into comparisons include LLAMA3, Sea Lion, and ChatGPT.
The innovation, which has been registered with the Ministry of Law, is now entering the clinical evaluation stage to test its implementation. In the future, LISA v2 will be equipped with an automatic satiation feature in order to increase user transparency and trust.