Application of deep learning-based facial pain recognition model for postoperative pain assessment
Authors' conclusion
The findings indicate that deep learning models leveraging facial expression analysis hold significant potential to recognize varying degrees of pain in clinical settings, especially severe pain.
Study context
- Population
- Postoperative patients (facial pain images) and volunteers (simulated pain images)
- Comparison
- Clinical Pain Dataset (CPD) vs Simulated Pain Dataset (SPD) vs Combined Dataset (CD)
- CPDn = 503
- SPDn = 51
- Outcome
- Model performance by dataset and pain level (AUROC and F1 scores)
Methodological detailsdeep learning model (pre-trained VGG16), clinical and simulated datasets, Facial pain recognition software developed based on the model
- Deep learning model (pre-trained VGG16)
- Clinical and simulated datasets
- Facial pain recognition software developed based on the model
Performance
As reported in the abstract
Severe pain identification (AUROC)
| Test or model | AUROC |
|---|---|
| CPD | 0.898 (95 %CI,0.877-0.917) |
| CD | 0.867 (95 %CI,0.844-0.889) |
AUROC
CPD0.898 (95 %CI,0.877-0.917)
CD0.867 (95 %CI,0.844-0.889)
Further findings
| Outcome | CPD | CD |
|---|---|---|
| Severe pain classification, training sets (AUROC) | ||
| AUROC (CPD-train) | 0.898 [95 % CI: 0.877-0.917] | — not stated in the abstract |
| AUROC (CD-train) | — not stated in the abstract | 0.917 [95 % CI: 0.883-0.948] |
— not stated in the abstract
Original publication
Application of deep learning-based facial pain recognition model for postoperative pain assessment. J Clin Anesth. Published 13 June 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →