Early Dementia Identification: On the Use of Random Handwriting Strokes

Vincenzo Gattulli, Donato Impedovo, Giuseppe Pirlo, Gianfranco Semeraro

Abstract: Timely diagnosis plays a crucial role for the treatment of neurodegenerative diseases. In particular, Dementia Identification in early stages is important to help patients have a better quality of life and to help clinicians to find a pathway of treatments to slow the effects. To the aim, a wide set of different handwriting tasks is here considered, and Shallow and Deep Learning methodologies are compared. Furthermore, Random Hybrid Stroke (RHS) are adopted to represent the handwriting time series. This solution outperforms the classical Deep Learning methodology and it is compared to a state-of-art shallow learning approach. Finally, a decision-level fusion for the results is adopted.