Professor Yu-Rang Park and researcher Jong-Hyeon Kim from the Department of Biomedical Systems Information at Yonsei University College of Medicine, Professor Gyeong-Soo Jeong and lecturer Min-Dong Seong from the Department of Respiratory Medicine at Severance Hospital and Dr. Hyun -Seok Min of Tomocube has developed an artificial intelligence model that can diagnose sepsis and predict prognosis.
Sepsis is a disease that causes damage to major organs due to the body’s abnormal response to infections and has a high rate of morbidity and mortality.
Because the immune response to sepsis is complex and varies from patient to patient, early diagnosis and timely action are important. Because it affects multiple organs rapidly, the chance of death increases if treatment is delayed.
Representative biomarkers currently used to diagnose sepsis, such as C-reactive protein (CRP) and procalcitonin (PCT), have delayed responses, resulting in a delay in diagnosis.
Furthermore, biomarkers such as interleukin-6 (IL-6), an inflammatory marker, lack standardization, making the interpretation of diagnostic results difficult. Due to these problems, it is necessary to discover new biomarkers.
The research team used immune cell CD8 T cell image data and an artificial intelligence model to determine whether the diagnosis and prognosis of sepsis could be predicted.
CD8 T cells were isolated from blood samples of 8 people in the sepsis recovery group and images were taken. Imaging was divided at the time of septic shock diagnosis (T1), resolution (T2), and before discharge (T3), and holotomography microscopy was used.
Holotomography technology can quickly and stably obtain 3D images of living immune cells without a staining process that affects changes in cell structure.
Images taken at each time point were compared and analyzed with images from 20 healthy control groups using a deep learning-based AI classification model.
The prediction performance of the AI model was analyzed using the “Receiver Operating Characteristic Curve (AUROC)” indicator.
AUROC means “area under the ROC curve” and is a statistical technique that indicates the diagnostic accuracy of a specific test instrument for diagnosing a certain disease. It is mainly used as an index to evaluate the performance of the AI model.
Typically, the closer it is to 1, the better the performance, and if it is 0.8 or higher, it is rated as a high-performance model.
As a result of the analysis, when only one cell image was used to diagnose sepsis, the prediction accuracy of the AI model (AUROC) was 0.96 (96%), while when two cell images were used, the performance was higher than 0.99 (99). %).
The prognosis prediction model also showed an accuracy of 0.98 (98%) using single-cell images and showed higher performance of over 0.99 (99%) when using two-cell images.
Professor Kyung-soo Jeong said: “Through this study, we were able to identify the role of 3D images of CD8 T cells as biomarkers of sepsis,” and added: “We expect that this will help make appropriate treatment decisions “.
#Development #artificial #intelligence #model #capable #diagnosing #predicting #prognosis #sepsis
