Deepnoid Takes Top Honors at KoSAIM 2026 with Seven Research Abstracts
Deepnoid secured the top oral presentation award and two poster prizes at the Korean Society of Artificial Intelligence in Medicine (KoSAIM) 2026 regular conference, which concluded on the 24th. The company presented seven research abstracts spanning chest imaging, brain MRI, and pathology during the event.
GLINT Vision-Language Technology Wins First Place in Oral Presentations
Their winning research centers on "GLINT," a medical imaging vision-language artificial intelligence technology that learns both the presence and location of clinical findings using only reading reports.
GLINT independently maps the correspondence between each sentence in a diagnostic report and its corresponding region within an image. It performs finding identification and location marking simultaneously on chest X-ray and computed tomography scans without requiring explicit location-labeling data. According to the company, the model outperformed existing models across comparative evaluation tasks and serves as a foundational architecture for various medical imaging AI workflows, including enhancing the diagnostic accuracy and reasoning disclosure capabilities of “M4CXR” and “M4CT.”
Two Research Teams Earn Poster Distinctions
In addition to the oral presentation win, two separate research efforts captured superior poster prizes at the conference. Lee Da-hye and Oh Kwan-seok received recognition for research utilizing foundation models to elevate the segmentation performance of white matter hyperintensity (WMH) lesions in brain MRI scans. Meanwhile, Cho Won-jun, Jung Dae-ki, Im Hyung-yeol, and Yoon Hong-jun earned poster honors for “STREAM,” a high-quality pathology image generation technology.
“AI that demonstrates its reasoning regions alongside its diagnostic decisions enables medical staff to utilize and trust the system with greater confidence,” stated researcher Lee Sung-eun. Lee added that the team intends to advance medical imaging foundation models and rationale-presentation technologies to integrate them directly into active clinical reading solutions.
Bridging Academic Validation and Clinical Workflow Integration
Following the conference awards, Choi Woo-shik stated an intention to connect validated academic research directly into services deployed in actual clinical reading environments.
