Seoul National University Researchers Develop Next-Generation Cancer Diagnostic Marker Applying 'AI Deep Learning Technology'

Graph-based artificial intelligence (AI) deep learning technology applied to cancer cell analysis. Provided by Seoul National University

[Asia Economy Reporter Yoo Byung-don] A research team at Seoul National University has discovered a new cancer diagnostic indicator using artificial intelligence (AI) deep learning technology.

On the 20th, Seoul National University announced that a joint research team led by Professor Kwon Seong-hoon of the Department of Electrical and Computer Engineering and Professors Moon Kyung-chul and Park Jeong-hwan of the Medical School developed deep learning technology that represents cancer tissue images as a 'cancer cell network' and provides diagnostic indicators in a form interpretable by medical professionals.

Recently, immunotherapy, which has attracted attention as a cancer treatment method, determines treatment success based on the 'cancer microenvironment,' the interaction between cells within the cancer tissue. Because of this, the cancer microenvironment itself is being utilized as a new cancer diagnostic indicator in clinical settings.

Until now, to use cancer microenvironment information as a diagnostic indicator, verification based on large amounts of data was necessary. However, AI could only learn and judge the local shapes of cancer cells and could not provide data interpretable by medical professionals in the field. Seoul National University stated that this limitation has been overcome through this research.

The research team explained, "By creating an AI that predicts the survival rate of cancer patients and interpreting its results, we were able to demonstrate that vascular formation within cancer tissue and the relationships between cancer cells and immune cells can serve as indicators for diagnosing survival rates."

They added, "The cancer cell network creation method and graph deep learning technology developed in this study are revolutionary approaches applicable not only to cancer tissue but also to any medical imaging data such as MRI and X-rays."

Meanwhile, the results of this study were published in the international academic journal Nature Biomedical Engineering.