CNN Based Analysis of the Luria’s Alternating Series Test for Parkinson’s Disease Diagnostics

Sergei Zarembo, Sven Nomm, Kadri Medijainen, Pille Taba, Aaro Toomela

Keywords: Parkinson's disease, drawing tests, data augmentation, CNN, diagnostics support

Abstract: Deep-learning based image classification is applied in this studies to the Luria's alternating series tests to support diagnostics of the Parkinson's disease. Luria's alternating series tests belong to the family of fine-motor drawing tests and been used in neurology and psychiatry for nearly a century. Introduction of the digital tables and later tablet PCs has allowed deviating from the classical paper and pen setting, and observe kinematic and pressure parameters describing the test. While such setting has led to a highly accurate machine learning models, the visual component of the tests is left unused. Namely, the shapes of the drawn lines are not used to classify the drawings, which eventually has caused the shift in the assessment paradigm from visual-based to the numeric parameters based. The approach proposed in this paper allows combining two assessment paradigms by augmenting initial drawings by the kinematic and pressure parameters. The paper demonstrates that the resulting network has the accuracy similar to those of human practitioner.

One-sentence Summary: Kinematic and pressure parameters have been used to enhance the drawings of the digital Luria's alternative series tests allowing the application of CCN architectures and as the tool to support diagnostics of the Parkinson's disease.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=IjnIMLjqwV