Table 2.
Features of wearable AI.
| Feature | Number of studies (%) | References |
|---|---|---|
| Wearable device | ||
| Actiwatch AW4 | 19 (35.2) | 16,18,26,27,29–31,35,36,40,41,52,55,57,59–62,69 |
| Fitbit series | 14 (25.9) | 20,21,25,28,33,34,42,44,46,47,50,53,63,68 |
| Empatica series | 3 (5.6) | 22,32,56 |
| Mi Band | 2 (3.7) | 19,45 |
| GENEActiv | 2 (3.7) | 43,49 |
| Others | 1 each (1.9) | 17,23,24,38,39,48,49,51,54,58,64–67 |
| Not reported | 1 (1.9) | 37 |
| Placement | ||
| Wrist | 50 (92.6) | 16–37,39–47,49–53,55–65,67–69 |
| Head | 1 (1.9) | 48 |
| Lower back | 1 (1.9) | 38 |
| Fingers | 1 (1.9) | 54 |
| Chest | 1 (1.9) | 66 |
| Waist | 1 (1.9) | 23 |
| Thigh | 1 (1.9) | 23 |
| Ankle | 1 (1.9) | 23 |
| Aim of AI algorithms | ||
| Detection | 48 (88.9) | 16–20,22–24,26–41,43–46,48–67,69 |
| Prediction | 6 (11.1) | 21,25,34,42,47,68 |
| Problem-solving approaches | ||
| Classification | 44 (81.5) | 16–26,28–31,33,34,36–43,45–49,51–55,57–62,66,68,69 |
| Regression | 5 (9.3) | 32,50,56,64,67 |
| Classification and regression | 5 (9.3) | 27,35,44,63,65 |
| AI Algorithms | ||
| Random Forest | 32 (59.3) | 16–19,21,22,25,28–34,36,39,42,43,45,46,48,50,51,54,55,57,59–62,64,69 |
| Logistic Regression | 13 (24.1) | 16,17,19,20,23,25,39,43,45,46,49,54,67 |
| Support Vector Machine | 11 (20.4) | 16,18,19,22,25,30,43,54,55,58,64 |
| Extreme Gradient Boosting | 10 (18.5) | 17,18,22,23,35,46,50,54,63,65 |
| Decision Tree | 8 (14.8) | 18,19,22,30,39,43,48,55 |
| AdaBoost | 8 (14.8) | 20,25,30,32,50,59,64,68 |
| Convolutional Neural Network | 6 (11.1) | 26,27,36,40,41,52 |
| Ensemble model | 6 (11.1) | 32,45–47,52,56 |
| K-Nearest Neighbours | 6 (11.1) | 17,20,22,30,54,55 |
| Long Short-Term Memory | 5 (9.3) | 24,37,38,40,41 |
| Gradient Boosting | 4 (7.4) | 17,20,22,64 |
| Multilayer Perceptron | 3 (5.6) | 22,23,66 |
| Artificial Neural Network | 3 (5.6) | 25,30,59 |
| Naive Bayes | 3 (5.6) | 30,48,55 |
| Gradient-Boosted Decision Trees | 2 (3.7) | 25,46 |
| Ridge Regression | 2 (3.8) | 32,44 |
| Gaussian Process | 2 (3.7) | 30,32 |
| Linear regression | 2 (3.7) | 32,67 |
| Deep Neural Network | 2 (3.7) | 31,36 |
| elasticNet | 2 (3.7) | 34,64 |
| Support Vector Classifier | 2 (3.7) | 17,23 |
| least Absolute Shrinkage and Selection Operator | 2 (3.7) | 20,44 |
| Others | 1 each (1.9) | 18,25,30,32,39,41,48,53,64 |
| Dataset source | ||
| Closed | 34 (63) | 19–22,24,25,28,30,32–34,37–39,42–51,53,54,56,58,63–68 |
| Open | 20 (37) | 16–18,23,26,27,29,31,35,36,40,41,52,55,57,59–62,69 |
| Data input to AI algorithm | ||
| Physical activity data | 47 (87) | 16–22,25–27,29–45,49–65,67–69 |
| Sleep data | 26 (48.1) | 19–21,25,28,32–34,39,42–47,49–51,53,54,56,63–65,67,68 |
| Heart rate data | 17 (31.5) | 17,19,21,22,24,25,42,44,45,50,51,53,54,56,64,65,67 |
| Mental health measures | 12 (22.2) | 21,25,32,34,39,44,46,47,49,51,54,64 |
| Smartphone usage data | 9 (16.7) | 19,20,32,50,51,54,56,67,68 |
| Location data | 9 (16.7) | 19,20,32,44,50,54,56,67,68 |
| Social interaction data | 8 (14.8) | 19,20,32,50,51,56,67,68 |
| Light exposure | 5 (9.3) | 21,39,42,51,65 |
| Demographic data | 5 (9.3) | 25,46,47,49,59 |
| Electrodermal activity data | 4 (7.4) | 17,22,32,56 |
| Circadian rhythms | 3 (5.6) | 23,49,63 |
| Skin temperature data | 2 (3.7) | 22,65 |
| Weather data | 2 (3.7) | 53,56 |
| Others | 1 each (1.9) | 32,37,43,46–48,51,53,64,66 |
| Ground truth assessment | ||
| MADRS | 19 (35.2) | 16,18,26,27,29–31,35,36,40,41,52,55,57,59–62,69 |
| PHQ-4, -8, and -9 | 14 (25.9) | 19,22,23,25,33,34,46–48,50,51,63,64,67 |
| DSM-IV and -5 | 5 (9.3) | 21,38,44,51,66 |
| HDRS | 5 (9.3) | 32,39,45,56,65 |
| BDI-II | 5 (9.3) | 17,20,37,51,68 |
| Clinical assessment | 2 (3.7) | 42,43 |
| STAI | 2 (3.7) | 24,37 |
| DASS | 2 (3.7) | 24,54 |
| DAMS | 2 (3.7) | 28,53 |
| QIDS | 2 (3.7) | 44,66 |
| GDS | 2 (3.7) | 22,39 |
| Others | 1 each (1.9) | 49,62,66 |
| Not reported | 1 (1.9) | 58 |
| Validation approach | ||
| K-fold cross-validation | 25 (46.3) | 17,19,22,25–27,29,30,32,33,38,40,45–47,51–53,56,59,62,63,65–67 |
| Hold-out cross-validation | 22 (40.7) | 21,23,24,26,27,29,32,37,39,41–43,45,51,52,56,58,60,61,66,68,69 |
| Leave-one-out cross-validation | 12 (22.2) | 18,20,27,28,31,32,35,36,44,48,50,68 |
| Nested cross-validation | 5 (9.3) | 16,34,54,57,64 |
| External validation | 2 (3.7) | 49,68 |
| Time-series cross-validation | 1 (1.9) | 54 |
| Not reported | 1 (1.9) | 55 |
BDI-II Beck Depression Inventory-II, DASS Depression, Anxiety and Stress Scale, DSM Diagnostic and Statistical Manual of Mental Health, HDRS Hamilton Depression Rating Scale, MADRS Montgomery-Asberg Depression Rating Scale, PHQ-9 Patient Health Questionnaire-9, QIDS Quick Inventory of Depressive Symptomatology, STAI State-Trait-Anxiety-Inventory.