Multi-class classification model using multi-layer hierarchical classifier. In hierarchical classifier, the first layer is used to distinguish control vs moderate ET and ...
Abstract:
Essential tremor (ET) is diagnosed and monitored by movement disorder specialists based on clinical observations. While many ET cases are benign, some require pharmacolog...Show MoreMetadata
Abstract:
Essential tremor (ET) is diagnosed and monitored by movement disorder specialists based on clinical observations. While many ET cases are benign, some require pharmacological and surgical management, and there is a need for tools to assist clinicians in making informed decisions. This work aimed to develop a computerized technique to detect the presence and severity of ET. A set of 6 writing and sketching tasks were performed by 39 subjects on a digital tablet. The position and pressure of contact during the sketching were recorded and analyzed to obtain the dynamics of drawing. ET patients were scored on the Fahn-Tolosa-Marin Tremor Rating Scale by blinded movement disorder neurologists, and then separated into two groups: moderate and severe ET. Drawing tasks were more effective than writing tasks in distinguishing the groups, with drawing horizontal and vertical lines being the most sensitive. A new set of composite index feature was found to be most suitable in separating the three groups, with a Spearman correlation coefficient of 0.72. The technique shows significant differences between controls, patients with moderate tremor and those with severe tremor, with an accuracy of 87.2%. Our computerized analysis significantly outperformed non-specialist clinicians in differentiating ET from control. We conclude that computerized analysis of the dynamics of sketching horizontal and vertical lines is a suitable method to assess the presence and severity of ET.
Multi-class classification model using multi-layer hierarchical classifier. In hierarchical classifier, the first layer is used to distinguish control vs moderate ET and ...
Published in: IEEE Access ( Volume: 9)