Table 6 Comparative analysis of related works with the proposed model based on various handwritten datasets for PD prediction.
Related works | Type of handwritten dataset | Dataset size | Models applied | Accuracy (%) |
|---|---|---|---|---|
Parkinson’s disease handwriting database (PaHaW) (spiral images) | 75 | Ensemble classifier | 74.76 | |
Real-time dataset (word, syllable, or sentence) | 75 | SVM, Adaboost, KNN | 81.3 | |
NewHandPD dataset (spiral and meander images) | 92 | CNN | 95.83 | |
Not mentioned (handwriting dynamics) | 66 | SVM-RBF | 83 | |
Parkinson’s disease spiral drawings using digitized graphics tablet dataset (spiral images) | 77 | CNN | 96.5 | |
Publicly available UCI ML repository (spiral images) | 80 | Adaboost | 96.02 | |
Innovation and technology assessment of the Federal University (spiral and wave images) | 204 | DenseNet201 and VGG16 | 94% and 90% | |
Publicly available dataset (spiral and wave images) | 469 | VGG19-INC | 98.45 | |
PaHaW dataset (spiral images) | 166 | CC-Net | 89.3 | |
Publicly available dataset (spiral and wave images) | 204 | CNN | 93.3 | |
Publicly available dataset (spiral and wave images) | 960 | RF, KNN, SVM | 83.1 | |
NewHandPD dataset (circle, meander, and spiral images) | 279 | ResNet, VGG19, InceptionV3, and KNN | 95 | |
Publicly available UCI ML repository (spiral images) | 77 | Restricted Boltzmann machine (RBM) pipelined with multi-layer perceptron model classifier | 95.32 | |
K Scott Mader dataset (spiral and wave images) | 3264 | Modified Mobile Net V2 | 97.7 | |
K Scott Mader dataset-augmented (spiral and wave images) | 3264 | EfficientNetB2 | 96.4 | |
RGG-Net | K Scott Mader dataset-augmented (spiral and wave images) | 3264 | ResNet-50 and GoogLeNet | 99.12 |
