JLK Secures Patent for "Weakly Supervised Learning-Based AI Technology"

Medical artificial intelligence (AI) company JLK (CEO Kim Dongmin) announced on November 26 that it has secured a patent for a technology that automatically analyzes the differentiation grade of gastric adenocarcinoma tissue by applying a deep learning-based "weakly supervised learning" method.

The core value of this patent lies in securing weakly supervised learning technology, which solves the problem of building big data for training-a process that is the most costly part of medical AI development.

Traditional medical AI development requires the creation of training data by "manually annotating" thousands to tens of thousands of image files at the pixel level. In particular, pathology images are large and structurally complex, making annotation extremely challenging and expensive-an industry-wide problem. However, JLK's patented technology presents a new way to address these high-cost and high-difficulty challenges through technology.

The core of the patented technology is a weakly supervised learning structure that enables training using only whole-slide information and partial labels, without the need for complex pixel-level annotation. In other words, even if humans do not manually indicate the precise location of cancer cells, the AI can independently identify major tissue regions and directly infer tissue grades such as well, moderately, and poorly differentiated. This significantly reduces both the cost and time required to build datasets compared to conventional methods, and enables the acquisition of larger datasets within the same budget, laying the groundwork for further advancement of AI.

This technology is not limited to a specific type of cancer; it is a general-purpose technology that can train on large-scale data with minimal labeling, and is thus considered a core technology that can enhance the competitiveness of JLK’s entire pipeline. JLK’s stroke AI solution (CT/MRI-based prediction and diagnosis), which is currently under intensive development, also requires multi-institutional, large-scale data. Therefore, weakly supervised learning-based data processing technology is highly significant in the stroke domain for data acquisition, learning speed, and the advancement of predictive models.

In the field of gastric cancer, this patented technology can be implemented in concrete applications. The AI automatically detects tissue regions in whole-slide images, then classifies the differentiation grades-well, moderately, or poorly differentiated-based on cancer cell differentiation, and quantitatively calculates the respective proportions (%) to provide to medical professionals. In particular, poorly differentiated tissue is closely associated with a high risk of lymph node metastasis and is critical for treatment decisions. Since the AI can automatically visualize and quantify poorly differentiated regions, the technology is seen as having strong potential for expansion into various clinical studies, such as predicting metastasis risk or determining treatment priorities.

JLK plans to further enhance data construction efficiency and technical scalability by expanding the application of this weakly supervised learning-based patented technology not only to gastric cancer pathology but also to a wide range of medical AI product lines, including stroke, dementia, cardiovascular diseases, and other pathology images. In particular, the company intends to accelerate its global strategy focusing on countries with a clear growth trend in the digital pathology and medical imaging markets, such as South Korea, the United States, and Japan.

Kim Dongmin, CEO of JLK, said, "This patent is more than just securing gastric cancer analysis technology-it is a strategic achievement that addresses the biggest barrier in medical AI development: the cost of data construction. The weakly supervised learning-based core technology will be a key asset for JLK to strengthen competitiveness across all business areas, including stroke, and to accelerate growth in the global market."