Movement disorders are a subset of neurological conditions that are responsible for a significant decline in the health of the world’s population, having multiple negative impacts on the lives of patients, their families, societies and countries’ economy. Parkinson’s disease (PD), the most common of all movement disorders, remains idiopathic (of unknown cause), is incurable, and without any confirmed pathological marker that can be extracted from living patients. As a degenerative condition, early and accurate diagnosis is critical for effective disease management in order to preserve a good quality of life. It also requires an in-depth understanding of clinical symptoms to differentiate the disease from other movement disorders. Unfortunately, clinical diagnosis of PD and other movement disorders is subject to the subjective interpretation of clinicians, resulting in a high rate of misdiagnosis of up to 25%. However, computerised methods can support clinical diagnosis through objective assessment. The major focus of this study is to investigate the use of machine learning approaches, specifically evolutionary algorithms, to diagnose, differentiate and characterise different movement disorders, namely PD, Huntington disease (HD) and Essential Tremor (ET). In the first study, movement features of three standard motor tasks from Unified Parkinson’s Disease Rating Scale (UPDRS), finger tapping, hand opening-closing and hand pronation-supination, were used to evolve the high-performance classifiers. The results obtained for these conditions are encouraging, showing differences between the groups of healthy controls, PD, HD and ET patients. Findings on the most discriminating features of the best classifiers provide insight into different characteristics of the neurological disorders under consideration. The same algorithm has also been applied in the second study on Dystonia patients. A differential classification between Organic Dystonia and Functional Dystonia patients is less convincing, but positive enough to recommend future studies.
