Deepnoid, a leading first-generation medical artificial intelligence (AI) company in Korea, announced on June 10 that three research abstracts evaluating the performance of its chest X-ray (CXR) report draft generation model, 'M4CXR,' have been accepted at ESTI 2025.
ESTI is an international radiology conference organized by the European Society of Thoracic Imaging. It is one of the most influential conferences worldwide and will be held in Bergen, Norway, from June 12 to June 14.
The research team led by Professor Jung Yeonju of Pusan National University Hospital used Deepnoid's M4CXR model to: ▲ assess its clinical applicability as a diagnostic support tool by evaluating the performance of report draft generation ▲ compare AI diagnostic performance using a multi-view learning approach ▲ and compare its performance with existing diagnostic support AI models. The accepted abstracts will be presented through oral presentations and poster sessions during the conference.
Professor Jung Yeonju, who will give the oral presentation, introduced the M4CXR model by stating, "It demonstrated an acceptable level of diagnostic accuracy for key radiologic findings such as nodules, ground-glass opacity, lung consolidation, emphysema, atelectasis, and pleural effusion." She added, "This suggests the potential for generative AI-based models to be used as diagnostic support tools in specific clinical settings."
Cho Honggeun, senior researcher at Deepnoid AI Research Institute, explained, "While existing chest imaging diagnostic support AI models are limited to about 5 to 10 major findings, M4CXR is differentiated by its ability to automatically generate report drafts for more than 40 types of findings." He also stated, "We expect this will enable more accurate diagnoses and contribute to improving the efficiency of clinical workflows. Follow-up studies will continue to explore its applicability to various lesions."
This year, Deepnoid developed the generative AI-based chest X-ray diagnostic and reporting model 'M4CXR' and is currently undergoing clinical trial procedures with the Ministry of Food and Drug Safety. Deepnoid aims to go beyond existing image data-based AI interpretation solutions and launch Korea's first and highest-performing generative AI-based medical device.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
