SEOUL, December 04 (AJP) - LG AI Research has developed a new pathology foundation model called “EXAONEPath” on Amazon Web Services (AWS) cloud infrastructure, capable of analyzing cancer tissue pathology images and reducing genetic testing time from two weeks to under one minute.
The model, unveiled at AWS re:Invent 2024 in Las Vegas, achieved 86.1 percent average accuracy across six benchmarks for image patch classification. Built on AWS's SageMaker inference chip and Amazon FSx storage system, it is part of LG's multimodal large language model EXAONE.
Using Amazon SageMaker, LG AI Research trained and deployed the EXAONEPath model in eight months, processing 285 million data points and over 35,000 high-resolution tissue sample images. The institute reduced data management and infrastructure costs by 35 percent and cut data preparation time by 95 percent.
The development leveraged AWS's cloud capabilities to transfer terabyte-scale data within an hour, shortening model training time from 60 days to one week. This advancement aims to help healthcare providers improve cancer diagnosis, reduce waiting times, and deliver personalized treatment options.
Copyright ⓒ Aju Press All rights reserved.