The Korea Institute of Machinery and Materials (President Park Sang-jin), under the Ministry of Science and ICT, has developed a technology that improves both the speed and accuracy of disease diagnosis by applying machine learning technology to medical imaging diagnostic equipment.

The research team led by Director Jong-Won Park of the Reliability Assessment Laboratory at the Korea Institute of Machinery and Materials announced on the 9th that they developed an “image diagnosis technology utilizing machine learning” by applying big data deep learning technology, which has been used to diagnose the reliability of machinery parts and equipment, to ultrasound imaging diagnostic equipment, and succeeded in diagnosing with an accuracy level of 80% using a GPU (graphics processing unit).

The research team has been seeking collaborations with medical researchers, who possess a wealth of image data, to develop machine learning techniques for testing the reliability of mechanical components and equipment. Recently, the medical field has also been actively applying machine learning techniques based on diagnostic imaging, such as ultrasound, computed tomography, and magnetic resonance imaging, to facilitate the early diagnosis of heart and brain diseases.

The research team, together with researchers from the Department of Cardiology at Daejeon St. Mary's Hospital, who are interested in the development of medical imaging diagnostic technology, began image analysis to diagnose aortic atherosclerosis in patients with cerebral infarction. While there have been various attempts to apply machine learning to the medical field, the development of a deep learning model that can be applied to classify aortic plaque status and measure plaque thickness is a novel endeavor.

The research team applied various machine learning techniques, including the autoencoder and U-net models, to the identification of ultrasound images of the aortic wall. Identifying the aortic wall using ultrasound images can help identify the condition of atherosclerotic plaques in the aorta, a known cause of stroke.

Park Jong-won, head of the Reliability Assessment Lab at the Korea Institute of Machinery and Materials, said, “Until now, users had to have complex data interpretation skills to determine the failure and lifespan of machinery parts and equipment, but now they can easily access it by utilizing various open sources.” He added, “Image diagnosis technology utilizing machine learning is expected to be utilized in various fields in the future, such as diagnosing various diseases and developing models to predict the lifespan of parts and equipment.”

The research team plans to improve the deep learning model to improve the accuracy of aortic plaque analysis. Furthermore, beyond the medical field, the team plans to expand the technology to include the construction of a virtual engineering platform for manufacturing future transportation equipment components and the Materials and Components Convergence Alliance (machinery and automotive) project, utilizing image data on component failures for fault diagnosis.

This research was conducted with the support of the Materials and Components Technology Innovation Project of the Ministry of Trade, Industry and Energy.