Researchers from the Seoul National University Bundang Hospital (SNUBH) have developed an artificial intelligence (AI) model to predict postoperative pain based on patients' facial expressions.

Accurately assessing pain is crucial for patient safety and recovery, and postoperative pain affects up to 71 percent of surgical patients and is highly subjective, making it challenging to evaluate, particularly for those who cannot communicate their pain effectively, such as children or individuals with certain mental health conditions.

The research team, led by Professors Koo Bon-wook and Park In-sun of the Department of Anesthesiology and Pain Medicine, aimed to create a quick and objective AI-based pain assessment model using facial expressions and physiological signals.

The study involved patients undergoing gastrectomy under general anesthesia, capturing facial expressions at different stages -- before surgery, immediately after surgery in the recovery room, during instances of pain requiring analgesics, and after pain relief. Additionally, they recorded the Analgesia Nociception Index (ANI), vital signs, and Numeric Rating Scale (NRS) pain scores.

The AI model, trained solely on facial expression data, predicted severe postoperative pain with high accuracy, surpassing models based on physiological signals such as ANI and vital signs.

The facial expression-based model achieved an Area Under the Receiver Operating Characteristic (AUROC) score of 0.93, indicating strong predictive performance.

In comparison, the combined model using facial expressions and vital signs data scored an AUROC of 0.84. An AUROC closer to 1 denotes better model performance.

"If we utilize artificial intelligence to quickly and accurately assess patients' pain in the post-anesthesia care unit, it is expected to contribute to the improvement of the quality of recovery for surgical patients through appropriate pain management treatment," Professor Koo said. "The model we have developed this time will be particularly helpful in assessing pain in patients who have difficulty communicating, as well as in post-surgical pain patients."

Professor Park also said, "Based on this research, our goal is to establish a system that can process large amounts of facial expression data from many patients using artificial intelligence so that medical staff do not have to individually assess patients' facial expressions and vital signs."

Through this, it will be possible to finely evaluate not only the presence of pain but also the intensity of the pain, Park added.

The study results were published in the Korean Journal of Anesthesiology.