This study investigates the classification of two prominent movement disorders: Parkinson’s Disease (PD) and Essential Tremor (ET) using a comprehensive machine learning framework. A novel dataset was meticulously created which contains handwriting samples collected at Habib Bourguiba Hospital in Sfax, Tunisia, specifically designed for differentiating between PD and ET. Preprocessing techniques such as image resizing, normalization, and data augmentation were employed to enhance robustness. Feature extraction was performed using the ResNet50 model, effectively capturing essential image characteristics through global average pooling. Recursive Feature Elimination (RFE) was then applied to identify the most significant features, followed by the training and validation of two classification models Random Forest and SVM using these selected features. The performance of these models is rigorously assessed through various metrics, revealing that the Random Forest model attained an accuracy of $92.83 \% \pm 2$, while the SVM model achieved an average accuracy of $94.66 \% \pm 1$. Visualizations such as confusion matrices and ROC curves provide deeper insights into model performance. Overall, the findings demonstrate the potential of machine learning techniques to enhance diagnostic accuracy in distinguishing between PD and ET, ultimately contributing to improved clinical decision-making.

Parkinson’s disease (PD) is a common and progressive neurodegenerative disorder with motor symptoms and a variety of non-motor symptoms. Experts regularly include handwriting as one of the Parkinsonian motor symptoms of PD and as a valuable tool that can aid in diagnosing and tracking the disease’s progression. PD patients have two periods. ‘On’ time is when levodopa is working well and your symptoms are controlled. ‘Off’ time is when levodopa is no longer working well and symptoms such as tremor, rigidity and slow movement re-emerge. To our knowledge, all existing publicly available datasets allow PD to be identified using only one period. No publicly available online handwriting datasets are dedicated to the analysis of PD using these two periods. Therefore, in this paper, we present our new online Arabic handwriting dataset for analysing PD, which we will make publicly available so that it could potentially be used for diagnosis, screening and monitoring the progression of PD. Our dataset was collected from 30 healthy controls and 30 PD patients in both “off” and “on” at the Neurology Department, Habib Bourguiba Hospital, Sfax, Tunisia. All participants performed five different handwriting tasks. The tasks included drawing repetitive ellipses, a spiral, repetitive digits and Arabic word writing. We hope that our new dataset will help researchers in the early detection of Parkinson’s disease, inpatient rehabilitation and quantification of therapeutic effects.

The high complexity of the brain, the large variation of tumors in size, shape and common textural features makes the manual Magnetic Resonance Imaging (MRI) analysis and interpretation, especially the classification of the tumor’ grades, a critical task during the clinical diagnosis. Therefore, an automated and highly precise tool is highly required for neuro-radiologists and neurosurgeons to differentiate and characterize the most frequent brain tumor ‘types.In this paper, we proposed an effective multi-classification framework to distinguish between three common types of brain tumors (“Glioma”, “Meningioma”, and “Pituitary”) from T1weighted contrast-enhanced–MR images. We investigated recent Deep Transfer Learning (DTL) technique and well-recognized pre-trained models on the ImageNet dataset basically the (DenseNet121, EfficientNetB0-V2, Xception, and Resnet50-V2).Using a publicly available benchmark dataset, the proposed framework achieved very satisfying promising results in term of accuracy and loss. the obtained accuracy over the testing dataset is 0.9635, 0.954, 0.9604, and 0.934 respectively for Xception, ResNet50-V2, EfficientNetB0-V2 and DenseNet121. For an effective evaluation, the performance of the studied models has been compared with Convolutional Neural Networks (CNN)’performances using the same dataset. The comparative study demonstrates that the selected networks outperforms the CNN in term of accuracy, specificity, and precision confirming therefore that the proposed framework could be considered as a second opinion routines for radiologist during the clinical diagnosis.

Classification of EEG signals based on motor imagery is an important task in Brain-Computer Interface (BCI). Deep learning approaches have been successfully used in several recent applications to learn features and classify different types of data. However, the number of researches using these approaches in BCI applications is very limited. In this paper, we aim at using the fusion of Convolutional Neural Networks (CNN) methods to improve the classification performance of EEG motor imagery signals in the framework of e-health Internet of Things. We propose and compare two classification methods based on the fusion of two CNNs. Our results show that the fusion of the CNNs with the Long Short-Term Memory (LSTM) layers offers a better classification performance compared to other state-of-the-art methods. The classification performance achieved by our proposed method using the BCI competition IV 2a dataset in terms of accuracy value is 61.68%. This method can be successfully applied to BCI systems where the amount of data is large due to daily recording.

Brain-Computer Interfaces (BCI) based on Motor Imagery (MI) extract commands in real time and can be used to control a cursor, wheelchair, robot or prosthesis by performing only mental imaging tasks, such as imagining a movement of the right hand while the corresponding brain activity is measured and processed by the system. Because MI-based BCI offers a high degree of freedom, it helps people with motor disabilities communicate with the device by performing a sequence of MI tasks. Several techniques are being developed to improve the classification performance of the MI signals used in BCI. Most researches focused on improving methods for feature extraction and selection, but relied on linear classifiers for class prediction. In this paper, we investigate the use of ensemble learning methods to improve classification accuracy in a BCI paradigm based on 2-class MIs. We propose and compare eight combinations of classifiers on the BCI Competition III dataset IVb. The results obtained show that the combination of three classifiers: Radial Basis Function-Kernel Support Vector Machine (RBF-Kernel SVM), Linear Support Vector Machine (Linear SVM) and Decision Tree, gives the best value of kappa which is equal to 0.783. This combination can be successfully applied to BCI systems where the amount of data stems largely from daily recording.