"Catching Cancer in 1 Minute"... LG AI Research Institute Develops Pathology Model Using AWS

'ExaOnePass' Development
Achieves Average Accuracy of 86.1%

Amazon Web Services (AWS) announced on the 3rd (local time) at 'AWS re:Invent 2024' held in Las Vegas, USA, that LG AI Research has developed a new pathology-based model (FM) for cancer diagnosis and treatment based on its cloud platform.

The new pathology-based model, 'EXAONEPath,' safely analyzes tissue pathology images of cancer patients and reduces the gene testing time from the existing two weeks to less than one minute, thereby helping medical staff improve the speed and effectiveness of treatment, AWS explained.

EXAONEPath achieved an average accuracy of 86.1% across six benchmarks related to the precise classification of image patches.

Through ExaOnePass, it is possible to analyze and visualize organizational pathology images to identify potential genetic variations within cells. Provided by LG AI Research Institute and AWS

EXAONEPath is part of LG AI Research's multimodal large-scale AI model 'EXAONE,' built on AWS's inference chip 'Amazon SageMaker' and the high-performance file and storage system 'Amazon FSx for Lustre.'

LG AI Research, LG Group's artificial intelligence (AI) research hub, transferred terabyte-scale data to the cloud within an hour, reducing model training time from 60 days to one week, thereby enhancing EXAONEPath's cancer diagnosis and detection performance, AWS reported. Additionally, by utilizing AWS, they reduced data management and infrastructure costs by about 35% and cut data preparation time by 95%.

Using Amazon SageMaker, LG AI Research trained and deployed the large-scale EXAONEPath model in eight months by utilizing 285 million data points and more than 35,000 high-resolution tissue sample images. AWS storage service 'Amazon S3' was used to store and retrieve the large-scale data essential for the research.

Lee Hwa-young, Executive Director of LG AI Research, said, "By leveraging AWS, we were able to train pathology models on vast datasets faster, more securely, and cost-effectively. The enhanced data processing capabilities of EXAONEPath will help provide more personalized and efficient cancer treatment, improving patient health."

Dan Siron, General Manager of AWS Healthcare and Life Sciences, said, "Through AWS, LG AI Research was able to develop and utilize EXAONEPath on an unprecedented scale, shortening data processing and model training times while improving accuracy. This will enable healthcare providers to improve cancer diagnosis and treatment, reduce waiting times, and offer patient-tailored therapies."