Polar Expression Feature of Digitized Handwritten Pattern for Automated- Parkinson’s-Disease Screening Using Perceptual Color Representation-Based Classifier

Abstract

Functional tremors are clear symptoms of neurodegenerative diseases; as such, they indicate the progression of Parkinson's disease (PD). Digitized handwritten pattern analysis of Archimedes' spirals, words, and sentences can help evaluate movement discords in the upper limbs. It offers a simple, comfortable, and repeatable method of examination for clinical applications and at-home monitoring usages. Upper limb tremors can be found in PD, essential tremor (ET), and cerebellar disorders. This paper proposes a quantitative method to scale the variations of functional tremors. The deviation (in cm) and the accumulation angle (in rad) of the feature pattern in polar expression were extracted to scale the variability at different tremor levels. Then, the proposed intelligent classifier, which is used as a perceptual color representation-based classifier (PCRC) and comprises a radial Bayesian network and a color relation analysis method, was employed to screen PD or ET with perceptual color representation. An assistant tool can integrate a smart mobile device (iPad/smartphone) and PCRC into the decision support system for individualized functions to evaluate the progression of the tremor level. The proposed decision support system was validated using data collected from 50 subjects. With fivefold cross-validation, average true positive, average true negative, and hit rates of 92.02%, 88.17% and 90.44%, respectively, were obtained to quantify the performance of the proposed classifier for identifying normal controls and PD or ET.

Keywords

• Parkinson’s diseases
• deviation
• accumulation angle
• radial Bayesian network
• perceptual color representation- based classifier