Classification of Handwritten Drawings of People with Parkinson’s Disease by Using Histograms of Oriented Gradients and the Random Forest Classifier

Abstract

Parkinson’s disease (PD) is a neurological disorder that is progressive and causes losses of dopaminergic neurons from the substantia nigra, a region in the human brain. The decrease of dopamine in this area elucidates the presence of motor symptoms, such as tremors, bradykinesia, rigidity, gait impairment, and non-motor symptoms, e.g., depression, loss of cognitive functions, sleep problems, and nerve pain. Among the motor symptoms, tremors can have the most impact on the social activities of people with PD. Furthermore, there is difficulty in diagnosing the underlying disorder that causes tremors. Thus, the study and development of methods to assess tremors and their severity is of paramount relevance for clinical practice. A typical clinical tool to evaluate tremor severity is the analysis of hand drawing shapes (e.g., spirals, circles, meanders, waves). The evaluation of these drawings is dependent on the experience of professionals, yielding a high variability of results. Aiming to contribute to the objective evaluation of hand drawing shapes of people with PD, this research proposes the application of the Random Forest Classifier to classify Histograms of Oriented Gradients (HOG) estimated from sinusoidal patterns collected from healthy individuals (n = 12) and from people with PD (n = 15). The highest accuracy, sensitivity and specificity classification success rates were of 83%, 85% and 81%, respectively. These results can be relevant for the early detection of pathological tremors, the follow-up of medical treatments and the diagnosis of parkinsonian conditions.