Developing an artificial intelligence model to predict postoperative pain with facial expressions
Jul 17, 2024
A research team led by Professor Koo Bon-wook and Park In-sun of the Department of Anesthesiology Pain Medicine at Seoul National University Bundang Hospital has developed an artificial intelligence model that predicts the occurrence of pain after surgery with only the facial expressions of patients and published research results confirming its usefulness.
Expressing pain is one of the important information reflecting the patient's health status, and proper evaluation and prompt response by medical staff are required for patient safety and rapid recovery. In particular, up to 71% of surgical patients are known to suffer after surgery, and the degree of pain is very subjective, and it is difficult to accurately predict the presence and intensity of pain when it is difficult to express their own pain, such as children or mentally ill patients.
Accordingly, a research team led by Professor Koo Bon-wook and Professor Park In-sun conducted a study to develop an artificial intelligence model that evaluates patients' pain after surgery using facial expressions and physiological signals that reflect pain.
For patients who underwent gastrectomy under general anesthesia, the research team photographed ▲ pain-free before surgery ▲ immediately after entering the anesthesia recovery room after surgery ▲ facial expressions with reduced pain after administration of painkillers. In addition, we measured the pain pain pain index (ANI), which is commonly used for pain monitoring, physiological signals such as vital signs, and the number pain scale (NRS), which expresses the patient's subjective pain intensity. Since then, an artificial intelligence model has been constructed by combining various collected data, and it has been verified whether it is possible to predict the intensity of pain after surgery.
The results show that AI models trained only on facial expression data predicted severe postoperative pain with very high accuracy, outperforming models based on physiological signals (pain sensation index, vital signs). Indeed, the prediction accuracy of AI models trained only on facial expressions was highest at AUROC 0.93, followed by models trained on facial expressions and vital signs data together (AUROC 0.84). AUROC is a performance indicator that represents the prediction accuracy of an artificial intelligence model, and the closer to 1, the better the performance.
Corresponding author Professor Koo Bon-wook "If artificial intelligence is used to quickly and accurately evaluate patients' pain in an anesthesia recovery room, it is expected that appropriate pain management treatment can contribute to improving the quality of recovery of surgical patients."The newly developed model will be of great help in evaluating pain not only for patients with postoperative pain, but also for patients with difficulty communicating."
Professor Park In-sun, the first author of this study, aims to build a system that can process a large amount of facial expression data of many patients using artificial intelligence without medical staff evaluating each patient's facial expressions and biological signals"This will allow a delicate assessment of not only the presence or absence of pain, but also the intensity of pain.'
Meanwhile, the results of this study were published in the Korean Journal of Anesthesiology, an SCIE-level international journal.
Expressing pain is one of the important information reflecting the patient's health status, and proper evaluation and prompt response by medical staff are required for patient safety and rapid recovery. In particular, up to 71% of surgical patients are known to suffer after surgery, and the degree of pain is very subjective, and it is difficult to accurately predict the presence and intensity of pain when it is difficult to express their own pain, such as children or mentally ill patients.
Accordingly, a research team led by Professor Koo Bon-wook and Professor Park In-sun conducted a study to develop an artificial intelligence model that evaluates patients' pain after surgery using facial expressions and physiological signals that reflect pain.
For patients who underwent gastrectomy under general anesthesia, the research team photographed ▲ pain-free before surgery ▲ immediately after entering the anesthesia recovery room after surgery ▲ facial expressions with reduced pain after administration of painkillers. In addition, we measured the pain pain pain index (ANI), which is commonly used for pain monitoring, physiological signals such as vital signs, and the number pain scale (NRS), which expresses the patient's subjective pain intensity. Since then, an artificial intelligence model has been constructed by combining various collected data, and it has been verified whether it is possible to predict the intensity of pain after surgery.
The results show that AI models trained only on facial expression data predicted severe postoperative pain with very high accuracy, outperforming models based on physiological signals (pain sensation index, vital signs). Indeed, the prediction accuracy of AI models trained only on facial expressions was highest at AUROC 0.93, followed by models trained on facial expressions and vital signs data together (AUROC 0.84). AUROC is a performance indicator that represents the prediction accuracy of an artificial intelligence model, and the closer to 1, the better the performance.
Corresponding author Professor Koo Bon-wook "If artificial intelligence is used to quickly and accurately evaluate patients' pain in an anesthesia recovery room, it is expected that appropriate pain management treatment can contribute to improving the quality of recovery of surgical patients."The newly developed model will be of great help in evaluating pain not only for patients with postoperative pain, but also for patients with difficulty communicating."
Professor Park In-sun, the first author of this study, aims to build a system that can process a large amount of facial expression data of many patients using artificial intelligence without medical staff evaluating each patient's facial expressions and biological signals"This will allow a delicate assessment of not only the presence or absence of pain, but also the intensity of pain.'
Meanwhile, the results of this study were published in the Korean Journal of Anesthesiology, an SCIE-level international journal.
|
bellho@sportschosun.com
