Table 4.
Summary of recent studies on automatic detection and staging of Parkinson's Disease.
| Selected study | Subjects | Data capture | Gait parameters and features | Detection algorithm | Result | Val. | ||
|---|---|---|---|---|---|---|---|---|
| Ricciardi et al. (2019) | 39PD and 7PSP | Mocap system | Spatiotemporal and kinematics | RF | ACC: 86.4% | 10-fold | ||
| Park et al. (2021) | 77PD and 34HC | Mocap system | Spatiotemporal and kinematics | RF | ACC: 98.1% | 5-fold | ||
| Ajay et al. (2018) | 16PD and 13HC | Vision-RGB | Spatiotemporal and kinematics | DT | ACC: 93.8% | 10-fold | ||
| PD Detection | Guayacán and Mart́ınez (2021) | 11PD and 11HC | Vision-RGB | Spatiotemporal saliency maps | 3D-CNN | ACC: 94.9% | LOSO | |
| Zhang et al. (2020a) | 656PD and 2148HC | IMU (smartphone) | Raw data augmentation | Ensemble of 5 CNNs | AUC: 0.86 | 5-fold | ||
| Zhao et al. (2018) | LSTM+CNN | ACC: 98.6% | 10-fold | |||||
| Xia et al. (2019) | CNN+Attn- BiLSTM | ACC: 99.1% | 5-fold | |||||
| El et al. (2020) | 1D-CNN | ACC: 98.7% | 10-fold | |||||
| Zeng et al. (2019) | RBF-NN | ACC: 98.8% | LOSO | |||||
| Selected study | Subjects | Data capture | Gait parameters and features | Detection algorithm | Stages | Result | Val. | |
| Lu et al. (2021) | 55PD | Vision-RGB | 3D human pose | CNN | MDS-UPDRS | ACC: 84.0% | LOSO | |
| Cao et al. (2021) | 18PD | Vision-RGB | Silhouettes | CNN | UPDRS | ACC: 84.2% | 3-fold | |
| Sabo et al. (2022) | 53PD | Vision-RGB | 2D human pose | GCN | UPDRS SAS |
F1: 0.53 F1: 0.40 |
LOSO | |
| PD Staging | Mirelman et al. (2021) | 332PD | 3-5 IMUs | RUSBoost | H&Y scale | AUC: 0.82 | 10-fold | |
| Veeraragavan et al. (2020) | ANN | H&Y scale | ACC: 87.1% | LOSO | ||||
| Alharthi et al. (2020) | CNN | H&Y scale | ACC: 95.5% | Hold out | ||||
| El et al. (2020) | 1D-CNN | UPDRS | ACC: 85.3% | 10-fold | ||||
| Balaji et al. (2021) | LSTM | UPDRS+H&Y | ACC: 96.6% | Hold out |
PD, Parkinson's Disease; HC, Healthy Control; PSP, Progressive Supranuclear Palsy; GRF, Ground Reaction Force; clfs., Classifiers; RF, Random Forest; CNN, Convolution Neural Network; ML, Machine Learning; RBF, Radial Basis Function; Attn, Attention-enhanced; (Bi)-LSTM, (Bidirectional) Long Short-Term Memory; ANN, Artificial Neural Networks; DT, Decision Tree; (MDS)-UDPRS, (Movement Disorder Society)-Unified Parkinson's Disease Rating Scale; SAS, Simpson-Angus Scale; H&Y, Hoehn and Yahr; ACC, Accuracy; AUC, Area Under Curve; F1, F1-score; LOSO, Leave-one-subject-out; Hold out, Random spilit.