Lunit Presents Two Medical AI-Focused Papers at International AI Conference 'CVPR 2023'

Medical AI company Lunit announced on the 12th that two regular papers were accepted at the AI conference 'CVPR (Computer Vision and Pattern Recognition) 2023.'

CVPR is a world-renowned conference in the field of computer vision and pattern recognition. At this year's CVPR, Lunit will present two papers specializing in medical AI technologies such as Pathology, beyond general computer vision (CV) related technologies.

The first paper introduces a new dataset called 'OCELOT.' OCELOT consists of data combining cell and tissue differentiation methods and was developed to improve cell detection performance in pathology images. Lunit expects OCELOT to enhance pathologists' cell detection capabilities as a more advanced pathology-specialized AI dataset.

The second paper presents research results that improve pathology image analysis performance by utilizing the latest technology called SSL (Self-Supervised Learning), which enables AI model training without annotations. Typically, expert annotations are considered essential for training precise AI models. However, this process requires considerable time and cost.

To overcome these drawbacks, Lunit conducted a benchmark study applying SSL technology to up to 33 million pathology images. As a result, they not only demonstrated the effectiveness of SSL but also achieved AI model performance improvements by leveraging Lunit's unique additional technologies.

Seobum Seok, CEO of Lunit, said, "This is the first time Lunit has presented computer vision papers specialized in medical fields at CVPR, proving world-class AI technology capabilities. The newly unveiled medical AI models will significantly contribute to the advancement of computational pathology technologies and provide innovative value to users."

Meanwhile, CVPR is an academic conference jointly hosted by the Institute of Electrical and Electronics Engineers (IEEE) and the Computer Vision Foundation (CVF) since 1983. This year, it will be held in Vancouver, Canada, from the 18th to the 22nd (local time).