Handwriting Dynamics as a Non-Invasive Digital Marker for Prodromal Alzheimer’s Disease

<title>Abstract</title> <p> <bold>Background:</bold> Handwriting impairment present a variety of characteristic involved with movement disorders and cognitive dysfunction in neurodegenerative diseases (NDDs). <bold>Objectives:</bold> Customized digital handwriting task (DHW) were performed to extract the characteristic profile for diagnosis of Alzheimer’s disease (AD) on early stage. <bold>Methods:</bold> Participants were recruited to perform whole DHW tasks, which included drawing a line, a cube, and writing a sentence in Chinese (Ch), English (E), and Korean (K), respectively. The AV45 PET-CT/MRI and p-tau217 in serum were used to AD diagnosis and evaluated the diagnostic performance of DHW. The attention-based one-dimensional convolutional neural network (1D-CNN) model was selected for DHW analysis. <bold>Results:</bold> A total of 252 NDDs patients and 185 age- and sex-matched controls were included, among who 61 participates accepted AV45 PET-CT/MRI and detection of p-tau217 in serum. The 1D-CNN model demonstrated that the DHW scores showed excellent discrimination between NDDs and controls, with an area under the receiver operating characteristic curve (AUC) of 0.961. In subgroup analyses, the cube task effectively distinguished AD from Parkinsonism, with AUC values of 0.832 which close to 0.857 in whole task. The diagnostic discriminative ability of AV45 ⁺ from AV45 ^- patients was evaluated, and the AUC for DHW and p-tau217 in serum were 0.702 and 0.732, respectively. The combination of DHW and p-tau217 had an increased AUC of 0.772. <bold>Conclusions:</bold> Customized digital DHW tasks can serve as a potential digital marker for AD screening on early stage. </p>