Deep Learning Vision Software for Doctors to Create Their Own Medical Image Analysis System

With the spread of COVID-19, the entire industrial landscape is changing dramatically. The medical industry is no exception. The unprecedented challenges posed by the pandemic is forcing the industry to make best use of scarce medical resources and make medical care more efficient by adopting innovative technologies.

Among various technologies applicable to the medical industry, deep learning vision technology (image interpretation using deep learning) is particularly receiving a lot of attention. The frequent use of CT, MRI, and X-ray scans in the medical sector entails advanced image interpretation technology. In this case, deep learning vision technology is an ideal solution given its ability to learn without much human supervision. However, practical limitations, such as the absence of deep learning experts and developers, input resources, and image security issues, thwart the active use of deep learning vision technologies in the medical field.

AI startup Neurocle is launching Neuro-T & Neuro-RT, a general-purpose deep learning vision software that can be individually adapted to the unique needs of the medical community while also offering an alternative to the barriers of traditional technology.

■ Deep learning with limited data and resources

There are already a variety of AI-based medical diagnostic solutions companies. But their focus is a disease-specific solution. To build such a solution, it is common for large hospitals and AI companies to sign MOUs, committing large amounts of data, labor, and money to provide a customized diagnostic service.

For rare diseases with limited data or new diseases in early stages of research, it is very difficult to find a suitable AI-based medical diagnostic solution. But with Neurocle's deep learning vision software, anyone can easily create a customized medical image analysis deep learning vision model.

△ Medical staff can create deep learning models that analyzes X-ray photos to aid diagnosis

■ Deep learning without deep learning experts

Information about deep learning vision technology is publicly available, but the usage and implementation of the technology requires deep learning experts and developers. For instance, when creating an image analysis algorithm that determines the presence of a disease, difficult concepts such as deep learning architectures and optimization of learning parameters must be considered to achieve the image analysis results the medical community needs.

This limitation can also be addressed by using software based on Auto Deep Learning that replaces the role of deep learning professionals. As long as you upload enough images to the software, the entire process of training a deep learning vision model can be done in a few clicks without coding. As the Auto Deep Learning technology automatically generates optimized deep learning models, it may be a particularly useful solution for medical personnel conducting real-world research.

■ Deep learning and information security

Despite the promising nature of web-based or cloud-based image analysis, security concerns precede the analysis and utilization of patients' medical images for research or diagnostic assistance. With the passage of the Data 3 Act, industry experts explain that while there is a growing expectation that pseudonymized data can be used for AI training, there is still a long way to go to freely utilize medical images.

If images can be analyzed and used directly on the user's local PC, this can settle the security concerns of the medical community. Neurocle is actively solving image security issues by providing an on-premises software that can be installed and used directly on PCs within hospitals or laboratories.

"In order to make deep learning applicable to a wide variety of disease images, we have made efforts to develop a deep learning vision software that anyone can use," said Neurocle CEO Hongsuk Lee, "Deep learning vision technology is no longer difficult to use, and it will be used more widely in the medical community in the future." In fact, Neurocle's deep learning vision software is currently being used by leading university hospitals in Korea to incorporate deep learning vision technology in their research.