A South Korean company has developed a platform that allows its users to easily perform AI-based medical research on a web browser without having to know anything about programming. This platform is an open platform that can be shared with researchers around the world and generate revenues.
DEEPNOID, a South Korean startup specializing in medical AI, is going to officially launch its medical AI research platform called “DEEP:PHI” that will help its users to develop complicated AI solutions without programming. With a single click, one can perform medical AI research based on deep learning by preprocessing 44 images that are already provided and assembling 24 neural network modules like blocks. This platform is perfect for hospitals, medical institutions, medical equipment makers, and research institutes that lack AI experts and expensive servers. DEEPNOID now allows its platform to be available for all medical teams when the platform was only limited to about 50 medical professionals who have agreements with the company.
For example, when data sets for learning are uploaded presuming that a model that classifies brain hemorrhage from CT scans of brain using AI is being developed, an icon is created on the project window. Project is completed by dragging and dropping image preprocessing modules such as size change and visibility improvement and pair neural network modules with the icon. When button is pressed, preprocessing and post-learning of uploaded CT scans of a brain take place. After testing the model with a new CT scan, results were 100% identical to answers.
AI research project can be completed in just couple minutes even if a user of the platform does not know anything about programming language. Senior researchers can check the internal structure made up of 152 layers and either revise codes or create brand new models. DEEP:PHI users will perform their research on a web browser while operations will take place within DEEPNOID’s server. Compression fraction diagnosis software called DEEP:SPINE that was created based on DEEP:PHI received an medical device approval from Ministry of Food and Drug Safety back in August and was verified of its effectiveness.
Researchers can decide whether to make their projects public or not. They can either perform private research for thesis or patent purpose or they can work together as the platform provides a function that allows them to perform joint research. DEEPNOID is also planning to launch “DEEP:STORE”, which is a marketplace for developed modules, at the end of April. This marketplace will help researchers to generate revenues through their projects.
“Unlike how most of medical AI companies provide illness deciphering solutions or platforms, we are the only company that provides a platform that helps researchers to perform entire process of their AI research without coding.” said Director Kim Tae-kyu of AI research institute who is also the co-founder of DEEPNOID. “DEEP:PHI provides tools for doctors who are not AI experts to lead research and it guarantees range of study and diversity of illnesses as it is an open platform allowing data sets to expand continuously.”
DEEPNOID is planning to have its platform available for the general public in the future. It is going to provide “DEEP:FOX”, which applies AI analysis solution to PACS (Picture Archiving and Communication System) in addition to DEEP:PHI and DEEP:STORE so that people can secure business continuity.
“Our goal is to help any researcher, medical professional, and data scientist who are starting AI research to easily research and use AI on a web browser.” said CEO Choi Woo-shik of DEEPNOID.
Staff Reporter Jung, Hyeonjung | iam@etnews.com