Cognitive Assessment through Writing Tasks

Casey Chen, Thomas Thebaud, Ankur Butala, Najim Dehak, Laureano Moro-Velázquez, Esther S. OhAlzheimer 2024 · DOI

TL;DR

This paper explores using handwriting analysis from digital tablet input to detect cognitive differences in individuals with Alzheimer's Disease (AD). By analyzing features like writing pressure and movement patterns during various tasks, researchers found significant distinctions between AD patients and healthy controls, suggesting handwriting analysis could be a valuable, non-invasive tool for AD assessment.

What Problem Does This Paper Address?

Alzheimer's Disease (AD) is a neurodegenerative condition that impacts not only memory but also motor and cognitive functions, including handwriting. While simple handwriting analysis has been used, it often misses subtle deficits. More complex tasks can reveal visual-spatial issues, but a standardized, sensitive method for early detection and assessment of AD through handwriting is still needed. This paper aims to bridge this gap by using advanced signal processing techniques to extract detailed, explainable features from handwriting data collected during various writing tasks, with the goal of improving the cognitive assessment of AD.

Methods

The researchers employed a digital tablet to collect handwriting data from participants. They designed three distinct writing tasks to probe different aspects of motor-cognitive function:

  • Copy Text: Participants copied a paragraph.
  • Copy Read Text: Participants copied a paragraph while also reading it aloud, introducing a multi-tasking challenge.
  • Freewrite: Participants wrote a paragraph spontaneously.

During these tasks, they recorded both on-tablet and in-air movements, as well as writing pressure. From this raw data, they extracted several "explainable features," including:

  • Pressure standard deviation (variability in applied pressure).
  • Writing time/completion time ratio (how much time was spent actively writing versus pausing).
  • Displacements between writing segments (both on-tablet and in-air).
  • Overall completion time.

Statistical analysis, specifically Welch's t-test with False Discovery Rate (FDR) correction, was used to compare these features between AD/MCI patients and healthy controls (HC).

Datasets

The study utilized a dataset of 60 participants:

  • 23 individuals diagnosed with Alzheimer's Disease (AD) or Mild Cognitive Impairment (MCI) due to AD.
  • 37 healthy controls (HC).

The data was collected using a digital tablet. This dataset appears to be newly collected for this study, as no mention is made of using existing benchmarks.

Key Insights and Ideas

  • Handwriting as a Window into Cognition: The study reinforces the idea that handwriting is a complex motor-cognitive act, and deviations in its execution can reflect underlying neurological changes.
  • Task Complexity Matters: Different writing tasks reveal different aspects of cognitive function. The multi-tasking nature of "Copy Read Text" and the open-endedness of "Freewrite" seem to be particularly effective in highlighting differences.
  • Beyond Simple Fluency: The focus on detailed features like pressure variation and in-air movements moves beyond basic handwriting fluidity to capture more nuanced motor control and spatial planning deficits.
  • Hesitation and Spatial Deficits: The findings suggest that AD patients exhibit a "hesitation effect" (longer in-air times, more in-air movement) and potential visual-spatial deficits, as evidenced by differences in horizontal displacement.

Key Results

The study found statistically significant differences between AD/MCI patients and healthy controls across several key features:

  • On-tablet time/total task time ratio: This was significantly different across all three writing tasks (p < 0.05), indicating AD patients spent proportionally less time actively writing on the tablet.
  • Pressure variation (Pressure Std): Significantly different for the Freewrite and Copy Read Text tasks (p < 0.001 for Freewrite, p = 0.019 for Copy Read Text). AD patients showed greater pressure variation.
  • Completion Time: A significant difference was observed for the Copy Read Text task (p = 0.0019).
  • In-air vs. On-tablet Behavior: AD patients spent more time in-air than on-tablet and exhibited greater writing pressure variation.
  • Freewrite Specifics: In-air displacements between writing segments were significantly different (p < 0.01). AD patients had longer average in-air time and smaller horizontal displacement during spontaneous writing.

Limitations and Future Work

Limitations: * The paper does not explicitly detail limitations, but the sample size of 60 participants (23 with AD/MCI) is relatively modest, which could impact the generalizability of findings. * The study focuses on AD and MCI due to AD; it's unclear how these findings would apply to other forms of dementia.

Future Work: The authors suggest that analysis of writing tasks with varying complexity is useful for assessing motor-cognitive abilities in AD patients. Future work could involve:

  • Expanding the dataset to include a larger and more diverse participant pool.
  • Investigating the predictive power of these handwriting features for disease progression.
  • Exploring the integration of these features with other diagnostic tools for a more comprehensive AD assessment.
  • Developing real-time algorithms for continuous monitoring and early detection.

Summary updated

Abstract

Abstract Background Alzheimer’s Disease (AD) is a neurodegenerative disease which affects motor‐cognitive abilities including handwriting. Past studies have used simple writing tasks to analyze handwriting fluidity, while more complex tasks can detect visual‐spatial deficits through spatial organization analysis. In this study, we extracted explainable features from the handwriting of AD patients performing various writing tasks using signal processing techniques and analyzed their importance in the cognitive assessment of AD. Method We collected a dataset from 60 participants, 23 with AD or MCI due to AD and 37 controls (HC), using a digital tablet. The writing tasks used involved various degrees of freedom: In Copy Text and Copy Read Text participants copied a paragraph, with and without reading it out loud, adding a multi‐tasking complexity layer, and in Freewrite participants wrote a paragraph spontaneously. In‐air and on‐tablet movements and writing pressure were collected. Multiple explainable features were extracted, such as pressure standard deviation, writing time/completion time ratio, displacements between writing, and completion time. Features from AD and HC participants were compared using Welch’s t‐test and adjusted using the false discovery rate (FDR) correction. Result The results show that the on‐tablet time/completion time ratio was significantly different between AD and HC (p‐value &lt; 0.05) across all three tasks, and pressure variation was significantly different between groups for the Freewrite and Copy Read Text tasks. For Copy Read Text, there was a significant difference between groups in completion time. AD patients spent more time in‐air than on‐tablet with greater writing pressure variation. For Freewrite, in‐air displacements between writing segments were significantly different (p‐value &lt; 0.01) between groups. AD patients had longer average in‐air time and smaller horizontal displacement. The results suggest a possible hesitation effect and a visual‐spatial deficit in AD patients. Conclusion Analysis of writing tasks with varying complexity proved useful in assessing the motor cognitive abilities of AD patients. Our results show statistically significant differences between AD and HC participants for the on‐tablet time/total task time ratio and in‐air compared to on‐tablet times across all three writing tasks suggesting greater hesitance in writing among AD patients.
BibTeX
@inproceedings{chen2024cognitive,
  author = {Casey Chen and Thomas Thebaud and Ankur Butala and Najim Dehak and Laureano Moro-Velázquez and Esther S. Oh},
  title = {{Cognitive Assessment through Writing Tasks}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.093186},
}
Figure