Table 6.
Comparison of methodologies and results of work oriented to the detection of FOG.
| Reference | Methodology | Results | |||
|---|---|---|---|---|---|
| Number of Patients/Episodes | Acquisition | Processing | Specificity (%) | Sensitivity (%) | |
| [14] | 5 | EEG* (cortical regions: frontal, central, parietal) | Time-frequency analysis with combinations of DWT* and SVM* | 89.5 | 83.1 |
| [51] | 20/98 | VR* and Acc* (hip) | Step rate, freezing and energy index | 84.1 | 70.1 |
| [52] | 20 | VR* and Acc* (hip) | Support vector machines, stride detection, spectral power and motor status threshold | 94 | 96 |
| [53] | 15 | VR* and Acc* (hip) | Support vector machines | >90 | >90 |
| [54] | 10/237 | VR* and Acc* (ankle, thigh and lower back) | Continuous wavelet transform | 81.01 | 84.9 |
| [55] | 18 | VR* and Acc* (hip) | Diffuse Logic: Freezing index, derived energy ratio, variation of the cadence and power spectrum | >86 | >78 |
| [56] | 18/>200 | Visual, motion and depth | Support Vector Machines and Logistic Regression classifier. | 91 | 91 |
| [57] | 10/237 | Acc* (ankle, thigh and lower back) | Power spectrum, Freezing index, FFT* | 81.6 | 73.1 |
| [58] | 8/237 | Acc* (ankle) | Classifier: Freezing index, energy, FFT* and statistical characteristics | 85 | 70 |
| [59] | 10/237 | 3 × Acc* | DL* (Convolutional Neural Networks) | 90.6 | 69.29 |
| [60] | 15/46 | Acc* and angular velocity (hip) | Automatic learning algorithm | 91.7 | 86 |
| [61] | 32 | IMU* sensor, Acc* of Smartphone (hip) | Variations of K during threshold crossings | 93.41 | 97.57 |
| [62] | 21 | IMU* sensor, (Acc*, gyroscope and magne-tometer) | Data representation + DL* (Convolutional Neural Networks) | 89.5 | 91.9 |
| [63] | 30/25 | VR* | Time-frequency analysis with combinations of FFT* and WT* | >95 | 75–83 |
| [64] | 7 | MEMS* (headset or shins) | Dynamic Time Warping and ANN* | 96.7 | 94.5 |
| [65] | 6 | EEG* (cortical regions: Frontal F4) | Short time Fourier Transform | 88 | 84.2 |
* EEG = Electroencephalography, DWT = Discrete Wavelet Transform, SVM = Support Vector Machine, VR = Video recording, Acc = Acceleration, FFT = Fast Fourier Transform, MEMS = Microelectromechanical systems, ANN = Artificial Neural Network, IMU = Inertial Measurement Unit, WT = Wavelet Transform, DL = Deep Learning.