Handwriting Analysis in Early-Dementia Screening: Experimental Design and Diagnostic Potential : A Comprehensive Review

Abstract

Traditional clinical assessments for Mild Cognitive Impairment (MCI) and dementia are time‑consuming and often lack the sensitivity required to detect early cognitive decline. Digitized handwriting and drawing have emerged as promising behavioral biomarkers, yet their clinical translation remains limited by fragmented protocols. This systematic review, conducted according to PRISMA guidelines, synthesizes 46 peer‑reviewed studies (2014–2025), mapping existing datasets and experimental protocols to describe the current methodological landscape in terms of task design, extracted features, and clinical interpretability. Results show a strong data polarization: 18 on 46 studies (39.1%) rely exclusively on the DARWIN dataset, focusing on Alzheimer’s disease (AD) vs. healthy controls, while 28 on 46 studies (60.9%) use independent datasets, 53.6% of which include MCI as a diagnostic target. Across protocols, two trends dominate: the widespread use of fine‑motor tasks and the increasing adoption of high‑executive‑load tasks, the latter particularly effective in distinguishing different stages of cognitive decline. Kinematic features - such as temporal regularity, movement smoothness, and in‑air trajectories - emerge as the most discriminative across studies reporting feature‑level results. Regarding computational performance, even if models reach accuracy values up to 97% for AD vs. controls, 47.8% present unclear analytical risk of bias due to small samples and limited reporting. In contrast, physiologically interpretable models report more conservative but clinically robust performance, ranging from 75.7% to 82.5%. Overall, this systematic review provides a rigorous cross‑disciplinary framework that bridges clinical validation and algorithmic optimization, offering the first integrated and up‑to‑date synthesis connecting clinical neuropsychology with computational modeling and supporting the development of reliable handwriting‑based biomarkers for early dementia diagnosis. © 2013 IEEE.