Streamlining Pediatric Vascular Exams with AI

When pediatric patients undergo surgery, medical teams must closely monitor their respiratory status. Among the various indicators, the carbon dioxide level in the blood plays a crucial role. During general anesthesia, obtaining accurate blood levels requires inserting a catheter into an artery to draw blood. However, this process is challenging for newborns and infants due to their delicate blood vessels. Repeated blood draws can lead to vessel damage or anemia, necessitating alternative methods.

Background of AI Model Development

Recently, South Korean researchers have paved the way for assessing a child's respiratory status using various vital signs measured during surgery, without the need for additional blood draws. A team led by Professor Kim Hyun-ho at St. Nicholas Children's Hospital, Seoul St. Mary's Hospital, Catholic University of Korea, announced the development of an artificial intelligence (AI) model to estimate blood carbon dioxide levels. This research analyzed data from 3,586 pediatric patients, achieving a significant reduction in average error by 23%.

End-Tidal CO2 Measurement and the Role of AI

Currently, a non-invasive method to monitor respiratory status is the 'end-tidal carbon dioxide' (ETCO2) measurement. This technique measures the carbon dioxide level in exhaled breath, but this value doesn't always perfectly correlate with the actual blood carbon dioxide levels. To reduce this discrepancy, the research team trained the AI with 10 pieces of information, including body temperature, ventilator settings, age, and disease information. The AI-estimated values showed an average difference of 2.73mmHg compared to actual blood test results, whereas the average error using only exhaled breath CO2 was 3.56mmHg.

AI Accuracy and Future Research Directions

This study further validated the AI's performance using independent patient data. Measurements from 50 pediatric patients at Chungnam National University Hospital showed an average error of 3.65mmHg, and similar results were observed with data from 499 patients collected at Seoul National University Hospital. However, it is still too early to definitively conclude that blood tests can be eliminated. The research team has not yet confirmed how much blood draws or arterial catheterizations would be reduced when the AI is used in actual operating rooms.

Professor Kim Hyun-ho emphasized, "While this model cannot completely replace arterial blood gas analysis, it can serve as a valuable 辅助工具 (auxiliary tool) in situations where arterial catheterization is difficult or frequent testing is challenging." He added, "Further research is needed to confirm its effectiveness through direct application on patients before it can be implemented in clinical practice."

Conclusion

The findings of this research have been published in the international journal 《Anesthesiology》, presenting new possibilities for the safe surgical care of pediatric patients. We look forward to further advancements in AI technology within the medical field, offering improved treatment environments for young patients.