AI Based Multi-modal Parameter of Peripheral Blood Cells (MMPBC) Predicts Survival Risk in Critically Ill Children

Overview

This study aims to investigate whether an AI prediction model based on blood cell multi-modal data can achieve early warning of survival risk in critically ill children through a large-scale multi-center cohort of critically ill children.

According to the definition of the United Nations Children's Fund (UNICEF), children are individuals between the ages of 0 and 18. Critically ill children are those who are admitted to the PICU or NICU and suffer from severe illnesses that require special treatment. These illnesses may endanger the child's life. Studies have reported that the international PICU mortality rate in developed countries is 2% to 3%; in recent years, the in-hospital mortality rate of PICU in China is 4.7% to 6.8%. The assessment of the survival risk of critically ill children has always been a focus of attention. Traditional assessment methods include physiological indicators, scoring tools, severity of illness, and diagnosis time, which can help doctors make decisions to a certain extent, but their predictive ability is limited and difficult to comprehensively reflect the child's physiological status and disease progression. With the development of technology and social progress, blood cell analysis is evolving towards a highly integrated platform of multiple cell analysis technologies that provide more accurate results, more comprehensive parameters, and faster detection. Cell analysis applications are increasingly focused on the identification and alarm capabilities of abnormal samples, including reticulocytes, nucleated red blood cells, and immature granulocytes. In 2009, Mindray Group, in collaboration with the National Key Laboratory of Fine Chemicals, developed a new nucleic acid-targeted fluorescent dye that meets the requirements of blood cell analysis (the patented fluorescent dye won the National Science and Technology Progress Second Prize). This breakthrough technology overcame international intellectual property barriers and developed the first high-end blood cell analyzer, the BC-6800, with functions to detect nucleated red blood cells and reticulocytes. The device has been successfully promoted to over 90% of tertiary hospitals in China. While detecting routine blood cell ratios, this blood cell analyzer actually generates a large amount of multi-modal data on cell distribution characteristics, including cell distribution width and abnormal cell ratios. However, so far, these multi-modal data have not been fully utilized in clinical practice. Preliminary exploration of multi-modal cell data has demonstrated its enormous value in predicting, diagnosing, and prognosing infectious diseases in small populations. This study aims to retrospectively collect clinical data and blood cell multi-modal data from NICU and PICU hospitalized children in multiple centers across China, to establish a national multi-center blood cell multi-modal database with no less than 100,000 people, and to use artificial intelligence technology to achieve accurate, repeatable, and unbiased identification and classification based on differences in cell morphology and structural distribution. A high-performance prediction model will be constructed in the discovery cohort to predict the survival risk of critically ill children; the performance of the model will be validated in the validation cohort to evaluate its applicability in the Chinese population of critically ill children. This study will provide solid evidence for evidence-based medicine based on multi-center cohort studies and offer potential new inspection technologies for predicting the survival risk of critically ill children, providing auxiliary decision support for clinicians, improving the survival rate of critically ill children, and promoting the development of precision medicine.

Conditions

Eligibility

Locations (1)

Outcomes

Central Contacts