Anesthesiology and pain medicine · 2026
Facial Expression-based Pain Detection: Development and Validation in the Post-anesthesia Care Unit.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Tarkhan Fazel, Bijani Ali, Alijanpour Shayan, Hamidi Seyed Hossein
Аннотация
BACKGROUND: Objective pain assessment in non-communicative patients remains challenging. Facial expressions are valid indicators of pain; however, manual coding is not practical.
OBJECTIVES: This study aimed to develop and validate a real-time facial pain assessment system (FPAS) for detecting pain using deep learning-based facial expression analysis.
METHODS: This cross-sectional study enrolled 50 postoperative patients in the post-anesthesia care unit (PACU) of a teaching hospital. The system performed parallel processing of facial action units using FaceMesh and emotion recognition using DeepFace, with individual baseline calibration and temporal stabilization. At 1, 5, 10, 15, and 30 minutes after PACU admission, the system generated a 0 - 10 Visual Analog Scale (VAS) pain score and a four-level categorical classification, which were simultaneously compared with patients' self-reported VAS ratings. Pearson correlation, the intraclass correlation coefficient (ICC), linear mixed-effects models, and classification metrics were used.
RESULTS: Fifty patients (60% female; mean age, 44.6 years) were enrolled. The Pearson correlation between the FPAS and patient self-reported VAS scores was 0.74 (P < 0.001). The ICC was 0.71 (95% CI: 0.64 - 0.77). Pain scores decreased significantly over time with both methods (P < 0.001), and the time × method interaction was not significant (P = 0.11). Categorical emotion classification showed 56% overall accuracy (Cohen's kappa = 0.43, P < 0.001), with the greatest agreement for the neutral class (sensitivity 72%, specificity 85%). For the four-level pain intensity classification, overall accuracy was 78.5%, with sensitivity ranging from 44% (severe pain) to 85% (no pain).
CONCLUSIONS: The automated FPAS demonstrated a strong correlation and acceptable agreement with patient self-reported VAS scores for postoperative pain monitoring in the PACU, supporting its potential as an adjunctive assessment tool. However, broader validation in larger, more diverse cohorts is needed before its clinical implementation.
OBJECTIVES: This study aimed to develop and validate a real-time facial pain assessment system (FPAS) for detecting pain using deep learning-based facial expression analysis.
METHODS: This cross-sectional study enrolled 50 postoperative patients in the post-anesthesia care unit (PACU) of a teaching hospital. The system performed parallel processing of facial action units using FaceMesh and emotion recognition using DeepFace, with individual baseline calibration and temporal stabilization. At 1, 5, 10, 15, and 30 minutes after PACU admission, the system generated a 0 - 10 Visual Analog Scale (VAS) pain score and a four-level categorical classification, which were simultaneously compared with patients' self-reported VAS ratings. Pearson correlation, the intraclass correlation coefficient (ICC), linear mixed-effects models, and classification metrics were used.
RESULTS: Fifty patients (60% female; mean age, 44.6 years) were enrolled. The Pearson correlation between the FPAS and patient self-reported VAS scores was 0.74 (P < 0.001). The ICC was 0.71 (95% CI: 0.64 - 0.77). Pain scores decreased significantly over time with both methods (P < 0.001), and the time × method interaction was not significant (P = 0.11). Categorical emotion classification showed 56% overall accuracy (Cohen's kappa = 0.43, P < 0.001), with the greatest agreement for the neutral class (sensitivity 72%, specificity 85%). For the four-level pain intensity classification, overall accuracy was 78.5%, with sensitivity ranging from 44% (severe pain) to 85% (no pain).
CONCLUSIONS: The automated FPAS demonstrated a strong correlation and acceptable agreement with patient self-reported VAS scores for postoperative pain monitoring in the PACU, supporting its potential as an adjunctive assessment tool. However, broader validation in larger, more diverse cohorts is needed before its clinical implementation.