Assessing parkinsonism & cerebellar dysfunction with spiral & line drawings

Abstract

Background and purpose: Recognising neurological diseases is challenging without accurate diagnostic tools. Therefore, many approaches have been taken to recognise and evaluate these diseases through speech, movement, or drawing modalities.The purpose of the study is to compare the recognition of Parkinson’s and cerebellar symptoms using spiral and line drawings recorded from the same subjects. We also investigate the importance of pin pressure in classification. Furthermore, an attempt is made to use the two types of drawings together for more accurate classification.

Methods: Images were generated from the raw data with and without pressure data. We then performed classification with the help of pre-trained and own deep learning feature extraction models. Mann-Whitney U test is used to test the significance of the results with a 0.05 significance level.

Results: The results showed that spiral drawings significantly performed better than lines (p-value: 0.001). Furthermore, combining the two types of drawings improves recognition when pressure is available (p-value: 0.017). However, no performance degradation can be expected without pressure data using one drawing task (p-value: 0.507).

Conclusion: The spiral is recommended as the primary drawing, but combining multiple drawings can contribute to a more confident recognition. By excluding pressure, no significant decrease is expected in the model’s performance.