Transformer Based Parkinson's Disease Classification from Hand Tremor Signal

Parkinson's disease (PD), which develops neurologically as a result of the destruction of nerve cells, affects most of the functions used in daily life such as speech, thinking, behaviour, and motor reflexes. Thus, the diagnosis of PD is not easy. Considering the history of the disease, the patient's complaints are evaluated by a specialist, and methods such as neurological examinations are used. Early diagnosis and classification of Parkinson's disease, which negatively affects daily life and has no definitive treatment, are aimed. In this study, deep learning techniques are used to contribute to the early and accurate diagnosis of the disease by using tremor signals. A deep neural network model was created to analyse the signal data using one-dimensional CNN and Transformer structures. The, model which was evaluated in three classes as low medium and high, achieved a test set accuracy value of 85.8%. The aim of the study is to explain the model that can be developed using deep learning techniques in the field of PD to both researchers and everyone interested in this field and to give possible topics to be studied in this field. Suggestions were made to improve the results of the study and for future studies.