JLK Publishes AI-Based Automatic Stroke Severity Analysis Paper: "Clinical Adoption Acceleration Expected to Drive Revenue Growth"

JLK, a company specializing in medical artificial intelligence (AI), announced on September 22 that it has published a research paper in the international journal Journal of Stroke on a deep learning algorithm that automatically classifies the severity of stroke in patients with atrial fibrillation. This achievement is expected to significantly increase its clinical utility while also having a positive impact on JLK's revenue growth.

Research teams from JLK, Chonnam National University Hospital, and Seoul National University Bundang Hospital developed a deep learning algorithm that automatically classifies the size of cerebral infarctions caused by atrial fibrillation as mild, moderate, or severe using diffusion-weighted MRI images (DWI) of stroke patients. The algorithm was trained on 1,091 DWI images collected from four hospitals between 2011 and 2021, and for external validation, it was tested on 1,265 DWI images collected from 11 different hospitals between 2017 and 2020.

The study found that the deep learning algorithm achieved a high percentage agreement of 87.4% with stroke experts in the external validation dataset. Another indicator of agreement, the Cohen's kappa value, was 0.81, surpassing the agreement between two stroke experts (74.6% percentage agreement, kappa value of 0.62).

Notably, the algorithm showed a significant association between the severity of cerebral infarction and the risk of symptomatic hemorrhagic transformation (sHT). Patients classified as severe by the algorithm had the highest incidence of sHT, while no sHT cases occurred among patients classified as mild. This suggests that the algorithm's classification results reflect actual clinical risk.

This study is the first case of developing a deep learning model that automatically classifies stroke severity based on DWI. Stroke severity is a crucial indicator for determining the timing of direct oral anticoagulant (DOAC) administration. Especially in environments with a shortage of specialists, this algorithm can assist less experienced physicians in deciding when to administer DOACs.

JLK highlighted the algorithm's rapid processing speed, requiring an average of about five seconds to generate the final classification result from the original DWI image. Such swift analysis improves efficiency not only in clinical research but also in real-world treatment settings, demonstrating its potential for commercialization. The company expects that the publication of this paper will contribute to the global adoption of the technology in clinical settings and the expansion of overseas exports, thereby driving substantial revenue growth.

Ryu Wiseon, Chief Medical Officer at JLK, stated, "This study is a result that can directly contribute to optimizing stroke patient treatment," adding, "We will expand the algorithm to various imaging modalities such as CT in the future, strengthen product competitiveness, and increase clinical utility to drive the company's growth."

This content was produced with the assistance of AI translation services.