Early diagnosis of Alzheimer's disease through handwriting analysis and deep learning: A review.

Clinical Snapshot

Key Findings

• P Value: Not reported at review level

• Effect Size: Narrative summary only: handwriting-based models generally achieve accuracy exceeding 80%; deep learning models (CNN, multimodal Transformer fusion) approach 90% accuracy in structured tasks (clock drawing, figure copying)

• Primary Outcome: Diagnostic accuracy of handwriting-based models for discriminating AD/MCI from healthy controls

• Nnt Or Sensitivity: Individual sensitivity and specificity values not pooled or summarised at review level; online kinematic markers (reduced velocity, prolonged in-air time, increased pausing, pressure instability) identified as recurring discriminative features across studies

• Confidence Interval: Not reported — no pooled estimates or confidence intervals provided

Clinical Application

Handwriting analysis using digital tablets is low-cost, non-invasive, and ecologically valid — theoretically feasible in primary care and community settings. However, standardisation of task paradigms, device specifications, preprocessing pipelines, and clinical thresholds remains unresolved. No commercially validated, regulatory-approved tool is described. Implementation would require prospective multicentre validation, clinical workflow integration studies, and health technology assessment. In Australia, early AD diagnosis is supported by the RACGP's cognitive impairment guidelines and the National Dementia Action Plan. Current TGA-approved diagnostic pathways rely on neuropsychological assessment, neuroimaging (MRI, FDG-PET), and CSF/blood biomarkers (including recently emerging plasma p-tau217 assays). Handwriting-based digital biomarkers are not currently part of any TGA-approved diagnostic device or PBS-listed assessment tool. The technology could complement existing RACGP-recommended cognitive screening tools (e.g., MMSE, MoCA, GPCOG) if prospectively validated in Australian populations, accounting for linguistic diversity, educational variation, and Indigenous community contexts. Funding pathways through the MBS for digital cognitive assessment tools would require MSAC evaluation. Research partnerships with Australian dementia research networks (e.g., AIBL, NHMRC-funded cohorts) would be needed to generate locally applicable validation data. Adults presenting with subjective cognitive concerns or early memory symptoms in primary care or memory clinic settings; potentially applicable to population-level cognitive screening programmes. Current evidence base is insufficient for individual diagnostic decision-making.

Abstract

Alzheimer's disease (AD) is one of the most prevalent neurodegenerative disorders worldwide, requiring early identification for timely intervention and to slow disease progression. However, existing diagnostic approaches, while effective at later stages, remain limited in detecting early-stage AD. Handwriting analysis has recently emerged as a non-invasive, cost-effective, and ecologically valid digital behavioral biomarker that reflects neurocognitive impairment. This review examines the role of handwriting as a neurocognitive marker for AD, focusing on integrating deep learning methodologies to enhance early diagnostic accuracy. It also elucidates the neurocognitive mechanisms linking handwriting behavior and AD, addressing current methodological and translational challenges. We performed a PRISMA-informed structured literature search and narrative synthesis of handwriting- and drawing-based studies for detecting AD/mild cognitive impairment (MCI), including offline handwriting images and online pen-stroke kinematics captured by digital devices. Task paradigms, data dimensions, preprocessing pipelines, modeling strategies (traditional machine learning and deep learning), evaluation practices, and translational considerations were summarized, and studies were organized by detection purpose and analytic approach. Our findings show that handwriting-based models generally discriminate AD/MCI from healthy controls with accuracy exceeding 80%, while deep learning models (e.g., convolutional neural network and multimodal Transformer fusion) approach 90% in structured tasks like clock drawing and figure copying. Online kinematic markers (e.g., reduced velocity, prolonged in-air time, increased pausing, and pressure instability) recur across studies, and multimodal integration with speech, gait, or facial signals can further improve sensitivity and ecological validity, although most studies are small and single-center.

References

1. 1.Tang, Q., Liu, J., Fan, C., Zhang, X., Zhang, C., & Qi, H. (2026). Early diagnosis of Alzheimer's disease through handwriting analysis and deep learning: A review. Journal of Alzheimer's Disease. Advance online publication. https://doi.org/10.1177/13872877261446573