Classification of the Tremor signal from Accelerometers and Gyroscopes in Multiple Sclerosis

Tremor, an involuntary rhythmic oscillatory movement of a body part is a common problem for patients suffering from multiple sclerosis (MS). This paper aims to use accelerometric and gyroscopic measurements of postural tremor from the upper limbs to determine whether a patient suffers from MS. The used data includes signals from a group of 16 MS patients (3 males and 13 females) and a group of 18 healthy control subjects (9 males and 9 females). Methods involving neural networks were used for signal classification from the power spectral density (PSD) of the given signals. Different fully-connected neural network (FNN), convolutional neural network (CNN) and recurrent neural network (RNN) architectures were explored. The best reached results were a recall of 100% and a precision of 89%, achieved by one of the proposed CNN models.