From: Classification of facial paralysis based on machine learning techniques
References | Objective | Facial movements | Ground truth | Tools | Dataset | Performance | Limitations |
---|---|---|---|---|---|---|---|
Chaoqun Jiang et al. 2020 [10] | FP classification (6 FP grades) | HB | LSCI scanners K-NN SVM NN | RGB images blood flow images 80 unilateral FP patients | Accuracy NN 96.77% K-NN 67.74% SVM 86.77% | ||
Xin Liu et al. 2020 [7] | FP classification (3 severity levels) | Rest Open mouth Closure the eyes lightly Elevation of eyebrows Pursing lips etc. | HB | PHCNN-LSTM | YouTube Facial Palsy Database Extended CohnKanade Database | Accuracy PHCNN-LSTM 0.9481% | Few public FP databases available Lack of various facial expressions in the datasets |
Jocelyn Barbosa et al. 2019 [6] | Health classification (normal/patient) FP classification (PP/CP) | Rest Raising of eyebrows Screwing-up of nose Smiling with showing of teeth | RLR RF SVM DT NB Hybrid | 440 2D images 60 normal subjects 40 PP patients 10 CP patients | Sensitivity RLR 85.9% RF 92.3% SVM 72.5% DT 90.2% NB 79.9% | No evaluation of FP degree No classification of facial paralysis grade Small dataset | |
Anping Song et al. 2018 [1] | FP classification (7 categories) | Rest Eye closed Eyebrows raised Cheeks puffed Grinning Nose Wrinkled Whistling | FNGS2.0 | IDFNP (Inception v3 CNN + DeepID CNN) | 2D images 860 FP patients | Accuracy 97.5% | |
Muhammad Sajid et al. 2018 [5] | FP classification (5 grades) | HB | CNNs GAN | 2D images 2000 Patients | Accuracy 92.60% | ||
Banita and Tanwar. 2018 [12] | Evaluation of FP 3 categories for patient (can be cured, cannot be cured, may or may not be cured) | HB | Fuzzy logic | 3D images 82 patients | |||
Ting Wang et al. 2015 [11] | FP classification (6 grades) | Raise eyebrows Close eyes Screw up nose Plump cheeks Open mouth | HB | FPASMs SVM (RBF Kernel) | 62 FP patients single-side and both-side | ||
Anguraj and Padma 2015 [13] | Classifying the severity of facial paralysis (normal–mild–moderate–severe) | Closing of eye Raising of eyebrows Opening of mouth Screwing of nose | SPSA FFBPN | 9 images (2D and grayscale) | Accuracy 94% Sensitivity 90% | 2D grayscale images Small number of images |