Parkinson’s disease (PD) is a chronic and progressive neurological illness affecting millions of people in the world. The cure for PD is not available. Drug therapies can handle some symptoms of disease like reducing tremor. PD is diagnosed with decrease in dopamine concentrations in the brain by using clinical tests. Early detection of the disease is important for the treatment. In this study, dynamic spiral test (DST) and static spiral test (SST) of PD patients were analyzed with pre-trained deep learning algorithms for early detection of PD. Fuzzy recurrence plot (FRP) technique was used to convert time-series signals to grayscale texture images. Several time-series signals were tested to observe the performances. The deep learning algorithms were employed as classifiers and feature extractors. Drawing and signal types’ performances for classifying PD were comprehensively investigated. In short, according to the experimental results Y signal produced the best results in DST approach and arithmetic combination of the Y and P signals performed better in SST method.

帕金森氏病（PD）是一种慢性和进行性神经系统疾病，影响世界上成千上万人。PD的治疗方法不可用。药物疗法可以处理某些疾病症状，例如减轻震颤。通过使用临床测试，PD被诊断为大脑中的多巴胺浓度降低。疾病的早​​期发现对于治疗很重要。在这项研究中，采用预先训练的深度学习算法对PD患者的动态螺旋测试（DST）和静态螺旋测试（SST）进行了分析，以早期发现PD。模糊递归图（FRP）技术用于将时间序列信号转换为灰度纹理图像。测试了几个时间序列信号以观察性能。深度学习算法被用作分类器和特征提取器。对绘画和信号类型对PD进行分类的性能进行了全面研究。简而言之，根据实验结果，Y信号在DST方法中产生了最好的结果，而Y和P信号的算术组合在SST方法中表现更好。