ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
ArticlePopulationPain SettingPain Scale (Ground Truth)ML ClassifiersOutputResultsAccuracy (DOP)Accuracy (PI)Average AccuracyAverage Accuracy
2
Artificial intelligence to evaluate postoperative pain based on facial expression recognition - Fontaine - 2022Adult patients from a single university hospitalPostoperative pain Numerical Rating Scale (NRS)Convolutional Neural Network (CNN)Pain Intensity estimationPain Intensity
- Accuracy = 53%

Detection of Pain
- Accuracy = 89%
89.00%53.00%85.05%73.90%
3
Deep Pain: Exploiting Long Short-Term Memory Networks for Facial Expression ClassificationUNBC-McMaster
database
Self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Convolutional Neural Network - Long short-term memory (CNN - LSTM)Pain Detection and Pain IntensityPain Detection
- Accuracy = 83%
- AUC = 93.3%

Pain Intensity Estimation
- MSE = 0.74
83.00%26.00%DOP = Detection of PainPI = Pain Intensity
4
Automatic Decoding of Facial Movements Reveals Deceptive Pain ExpressionsHealthy SubjectsCold pressor-induced painPain stimuli-dependent assessmentsSupport Vector Machine (SVM)Detection of genuine vs. faked painAUC = 0.91%
Accuracy = 85%
85.00%N/A
5
Automatic vs. Human Recognition of Pain Intensity from Facial Expression on the X-ITE Pain DatabaseX-ITE Pain DatabaseHeat-induced and electrical-induced painNRS categorized into 4 pain intensities (no pain, low, medium, and severe)2-CNN's with sample weighting (complementing each other)Pain Intensity estimation with two groupings of pain levels (none/low/severe & none/moderate/severe)Accuracy = 51.7%N/A51.70%
6
Ensemble neural network approach detecting pain intensity from facial expressions - ScienceDirectUNBC-McMaster databaseself-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Convolutional Neural Network - Recurrent neural network (CNN - RNN)Pain Intensity estimationUNBC-McMaster
- Accuracy = 86%
- AUC = 90.5%
N/A86.00%
7
Ensemble neural network approach detecting pain intensity from facial expressions - ScienceDirectMIntPAIN
database
electrical-induced painStimuli-based pain levels (0–4)Convolutional Neural Network - Recurrent neural network (CNN - RNN)Pain Intensity estimationMIntPAIN
- Accuracy = 92.26%
- AUC = 93.67%
N/A92.26%
8
A novel approach for pain intensity detection based on facial feature deformations - ScienceDirectUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Double Machine Learning (DML) combined with Support Vector Machine (SVM)Pain Intensity estimation (PSPI)Accuracy = 96%N/A96.00%
9
Automatically Detecting Pain in Video Through Facial Action Units | IEEE Journals & MagazineUNBC-McMaster databaseself-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Support Vector Machine (SVM)Pain detectionAccuracy = 80.9%
AUC = 84.7%
80.90%N/A
10
Enhanced deep learning algorithm development to detect pain intensity from facial expression images - ScienceDirectUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Hybrid Convolutional Neural Network (CNN) - bidirectional Long Short-term Memory (LSTM)Pain Intensity Estimation, categorized into four levels (PSPI 0, 1, 2-3, greater than 4)Accuracy = 85%
AUC = 88.7%
MSE = 0.21
MAE = 0.18
F-measure = 78.2%
85.00%79.00%
11
Automatic coding of facial expressions displayed during posed and genuine pain - ScienceDirectUniversity studentsCold pressor-induced painPain stimuli-dependent assessmentsGaussian Support Vector Machine (SVM)genuine vs. faked painAccuracy = 88%88.00%N/A
12
Automated detection of pain levels using deep feature extraction from shutter blinds-based dynamic-sized horizontal patches with facial images | Scientific ReportsUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)K-Nearest NeighborPain Intensity Estimation, categorized into four levels (PSPI 0, 1, 2-3, greater than 4)Accuracy = 95.57%
Average F1 = 95.67%

(The F1 score is calculated as the harmonic mean of the precision and recall scores)
95.57%
13
The modeling of human facial pain intensity based on Temporal Convolutional Networks trained with video frames in HSV color spaceUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Temporal Convolutional Network (TCN)Pain Intensity Estimation, categorized into four levels (PSPI 0, 1, 2-3, greater than 4)Accuracy = 94.14%
AUC = 91.3%
MSE = 0.186
MAE = 0.234
94.14%81.40%
14
The modeling of human facial pain intensity based on Temporal Convolutional Networks trained with video frames in HSV color spaceMIntPAIN
database
electrical-induced painStimuli-based pain levels (0–4)Temporal Convolutional Network (TCN)Pain Intensity Estimation, categorized into five levels (0-4)Accuracy = 89%
AUC = 92%
MSE = 0.22
MAE = 0.26
89.00%78.00%
15
Multiview Distance Metric Learning on facial feature descriptors for automatic pain intensity detection - ScienceDirectUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Support Vector Machine (SVM)Pain detection

Estimation of pain intensity, categorized into four levels: PSPI 0, 1, 2, and ≥3
Pain detection
Accuracy = 89.59%

Pain intensity estimation
Accuracy = 75%
89.59%75.00%
16
Metrological Characterization of a Pain Detection System Based on Transfer Entropy of Facial Landmarks | IEEE Journals & MagazineUNBC-McMaster
database
self-identified shoulder pain Visual Analog ScaleLinear discriminant analysisPain detection (VAS≥0)

Pain intensity (VAS) estimation
Pain detection
AUC = 0.87

Pain intensity estimation
MAE = 2.44
85.00%97.56%
17
Self-supervised pain intensity estimation from facial videos via statistical spatiotemporal distillation - ScienceDirectUNBC-McMaster
database
self-identified shoulder pain Prkachin and Solomon Pain Intensity (PSPI)Convolutional Neural Networks (CNNs)Estimation of pain intensity

UNBC-McMaster: 16 pain levels
Training with BioVid and testing on UNBC-McMaster

Self-supervised model: AUC = 0.692
Supervised model: AUC = 0.801
75.00%
18
Self-supervised pain intensity estimation from facial videos via statistical spatiotemporal distillation - ScienceDirectBioVid database (part A)heat-induced painstimuli-based pain (5 levels)Convolutional Neural Networks (CNNs)Estimation of pain intensity

BioVid: 5 pain levels
Training with UNBC-McMaster and testing on BioVid

(Self-supervised) model: AUC = 0.655
(Supervised) model: AUC = 0.755
75.00%
19
Automated Assessment of Children’s Postoperative Pain Using Computer Vision | PediatricsPediatric patients from a tertiary care centerpostoperative pain Numerical Rating Scale (NRS)Logistic regression and linear regression modelsPain Detection (NRS ≥ 4)

Pain intensity estimation (NRS)
Pain Detection
Baseline pain: AUC = 0.84

Pain intensity estimation
Baseline pain: r = 0.47; z = 4.4 *
85.00%47.00%
20
Deep Learning for Identification of Acute Illness and Facial Cues of Illnessvolunteersstates of acute illness (fake)use of lipopolysaccharide (LPS) which triggers an immune response resembling acute illness Neural transfer convolutional Neural Network (NT-CNN) and four Convolutional Neural Networks (CNN)Pain detectionPain Detection
AUC = 0.67
67.00%N/A
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100