AI model reduces cost and time, improves accuracy of cancer diagnosis

The research group of Park Sang-hyun (Head of Artificial Intelligence Major), Professor of the Department of Robotic Mechatronics Engineering at DGIST (President Guoyang), developed a "weakly supervised deep learning" model, which is based only on the data of the site where the cancer occurs.

Under normal circumstances, it is necessary to accurately mark the location of the cancerous site in order to solve the partitioning problem involved in marking the cancerous location information, which is time-consuming and cost-increasing

□ To address this problem, weakly supervised learning models are under active research, which use only rough data (such as "is the cancer present in the image") to segment the cancer site

□ In this regard, Professor Park Sang-hyun's research team discovered a technique for subdividing cancer cells into cancer sites based only on known slides showing the presence of cancer

□ The newly developed deep learning model achieves high Dice Similarity Coefficient (DSC) scores of 81 - 84 when using only the learned data with sliding horizontal cancer labels in the cancer partitioning problem

□Professor Park Sang-hyun said, "The model developed through this research has greatly improved the performance of weakly supervised learning of pathological images, and is expected to help improve the efficiency of various studies that require pathological image analysis

□ At the same time, the research results have been recognized for excellence and published in the most authoritative international academic journal "MediIA" (Medical Image Analysis Journal) in the field of medical image analysis