Dr. Se-Jong Kim and Dr. Juwon Na, leading the research team at the Materials Data Management Center within the Materials Digital Platform Division, in collaboration with Professor Seungchul Lee's team at POSTECH, have achieved a breakthrough in the field of materials science. Their innovation lies in the development of a technology capable of automatically identifying and quantifying materials' microstructures from microscopic images, using a unique blend of human-in-the-loop machine learning. Notably, their efforts have been supported by the Korea Institute of Materials Science (KIMS), a government-funded research institute under the Ministry of Science and ICT.

Figure 1. Microstructure Segmentation Based on Human-AI Collaboration. (Credit: Korea Institute of Materials Science (KIMS))

Figure 1 shows microstructure segmentation based on human-AI collaboration. Microscopic imaging systems offer insights into material structures spanning the nanoscale to the mesoscale. Quantitatively analyzing microstructures involves extracting statistical data from these microscopic images. However, due to the intricate and diverse nature of microstructures, both human and AI have faced significant challenges when attempting to perform this analysis in isolation.

In an approach, the research team has successfully integrated human expertise and AI capabilities to create a comprehensive framework for quantitative microstructure analysis. This technology empowers AI to undertake microstructure segmentation, utilizing just a single microstructure image and accompanying scribble annotations provided by domain experts. Moreover, AI actively engages with human experts, soliciting additional scribble annotations to enhance the model's performance and reliability. Extensive experimentation has confirmed the universal applicability of this human-AI collaboration framework, making it adaptable to a wide array of materials, microstructures, and microscopic imaging systems.

This innovative approach departs from previous research practices that necessitated extensive dense annotations, replacing them with easily drawn scribble annotations using a pen or mouse. The outcome of this research will be integrated into the Automated Microstructure Quantitative Analysis System (TIMs) currently under development at KIMS, facilitating its accessibility to researchers across various disciplines.

Dr. Juwon Na, a senior researcher at KIMS, remarked, "This study is the result of improving the existing subjective and time-consuming quantitative analysis of microstructure into an objective and automated process,” and Professor Seungchul Lee of POSTECH, added:“Our framework that interacts with experts is expected to be widely used as a core analysis technology in industry and research, and through this, we expect to dramatically reduce the cost and time of new materials research and development and further significantly improve reliability."

The research received support from the Ministry of Science and ICT through the basic project of the Korea Institute of Materials Science, the mid-career researcher support project of the National Research Foundation of Korea, and the Alchemist project of the Ministry of Trade, Industry, and Energy. The research findings were published on August 15 in Acta Materialia, the premier journal in the metallic materials field, with Dr. Juwon Na as the first author.

Source: National Research Council of Science & Technology

Cite this article:

Hana M (2023), Materials Microstructure Analysis: Swift and Simple with Human-AI Synergy!, AnaTechmaz, pp. 623