Dynamics of Handwriting for Cognitive Assessment

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

TL;DR

This study explores the potential of analyzing the dynamics of handwriting, specifically spiral drawing, as a non-invasive marker for Alzheimer's Disease (AD). Researchers found significant differences in drawing speed, duration, and frequency patterns between individuals with AD and healthy controls, suggesting that subtle motor changes captured by digital tablets could aid in early cognitive assessment.

What Problem Does This Paper Address?

Diagnosing Alzheimer's Disease (AD) often relies on cognitive tests and neuroimaging, which can be invasive, expensive, or not readily accessible. Previous research has hinted that motor dysfunction can be an early indicator of AD. However, there's a need for simpler, more accessible, and objective methods to detect these early signs. This paper investigates whether analyzing the dynamic characteristics of a simple handwriting task – drawing spirals – on a digital tablet can reveal subtle motor differences indicative of AD, going beyond just the visual appearance of the drawing.

Methods

The researchers collected handwriting data from 23 participants diagnosed with AD and 40 healthy controls (CTRL). Participants were asked to draw spirals in three ways: using their left hand, right hand, and dominant hand, while also repeating a specific syllable ("Pa-ta-ka"). The drawings were captured on a digital tablet, allowing for the analysis of not just the final shape but also the process of drawing.

From these digital recordings, 14 distinct features were extracted. These features captured time-independent aspects of the spirals and their drawing dynamics, including:

  • Duration: How long it took to draw the spiral.
  • Fundamental Frequencies (FF): Analyzed using Fourier Transform, these represent underlying rhythmic patterns in the drawing motion, measured from both the origin and center of the spiral.
  • Average Loop Speed: The speed at which loops within the spiral were drawn.

These extracted features were then statistically compared between the AD and CTRL groups using Welch's t-test, with a False Discovery Rate (FDR) correction applied to account for multiple comparisons.

Datasets

The study utilized data collected from 63 participants (23 AD, 40 CTRL) over multiple sessions. This dataset, comprising digital tablet handwriting samples of spiral drawings, appears to be a newly collected dataset for this specific research purpose, focusing on dynamic features rather than static visual analysis.

Key Insights and Ideas

  • Dynamics over Visuals: The most striking insight is that while the visual appearance of spirals drawn by AD patients and controls might not differ significantly, the way they are drawn (the dynamics) reveals crucial differences. This highlights the power of high-resolution digital data capture.
  • Subtle Motor Impairments: The study suggests that even in early stages, AD can manifest as subtle changes in motor control, affecting the smoothness, speed, and rhythm of handwriting.
  • Hand Dominance Matters: Differences were observed across all drawing conditions (left hand, right hand, dominant hand), but the magnitude and significance of these differences varied, indicating that hand dominance might play a role in how AD affects motor control.
  • Frequency as a Biomarker: The fundamental frequencies extracted from the spiral drawings appear to be particularly sensitive indicators, with higher frequencies observed in AD participants. This could relate to changes in tremor or oscillatory patterns during movement.

Key Results

The study found statistically significant differences between AD and CTRL groups across various spiral drawing tasks and features:

  • Right Spirals: Significant differences were found in Duration and Fundamental Frequency (FF) (p<0.01), and Average Loop Speed (p<0.05).
  • Left Spirals: All analyzed features showed significant differences (p<0.05).
  • Pataka Spirals: Significant differences were observed in Duration, Average Loop Speed, and FF (both Origin and Center) (p<0.05).
  • All Spirals (Combined): When all spiral samples were analyzed together based on group type (AD vs. CTRL), all features showed a significant difference (p<0.01).

In conclusion, AD participants tended to take longer to draw spirals, resulting in a lower average speed per loop compared to controls. Additionally, higher frequencies were detected in the drawing dynamics of AD participants.

Limitations and Future Work

  • Sample Size: While 63 participants is a reasonable starting point, a larger and more diverse cohort would strengthen the generalizability of the findings.
  • Visual vs. Dynamic: The paper explicitly states there was "no notable visual variation" between AD and CTRL spirals. This implies that traditional visual inspection might miss early signs, reinforcing the need for dynamic analysis. However, it also means the study didn't explore if specific visual features might also be affected.
  • Future Work: The authors suggest that these dynamic handwriting characteristics could serve as a valuable, non-invasive tool for cognitive assessment, potentially aiding in the early detection and monitoring of AD. Further research could explore the correlation of these features with disease progression and other cognitive measures. Investigating the specific neurological underpinnings of these observed motor differences would also be a valuable next step.

Summary updated

Abstract

Abstract Background Previous studies have shown that motor dysfunction can be an indicator of AD. We measured the observable and dynamic characteristics of spirals drawn by people with AD. Method We instructed 23 AD and 40 healthy control (CTRL) participants to perform a range of handwriting tasks on a digital tablet using a pen. We instructed the participants to draw spirals in three different ways with: i) left hand; ii) right hand; iii) dominant hand while repeating the occlusive word “Pa‐ta‐ka.” From the drawing samples of 63 participants collected over multiple sessions, we computed 14 time‐independent, explainable features from the shape of the spirals and the dynamics of the drawings (duration, fundamental frequencies using Fourier transform etc.). Those features were compared between the two studied groups using Welch's t‐test with a false discovery rate (FDR) correction. Result Right Spirals We found there is a significant difference (p‐value&lt;0.01) for the Duration and Fundamental Frequency (FF) with respect to Origin vs Total Number of Points, and FF with respect to Center vs Total Number of Points. Furthermore, the Average Loop Speed is significant (p‐value&lt;0.05). Left Spirals We found there is a significant difference (p‐value&lt;0.05) for all features. Pataka Spirals Additionally, there is a significant difference (p‐value&lt;0.05) for AD and CTRL participants concerning the Duration, Average Loop Speed, and FF Origin and FF Center. All Spirals We compared all spiral samples based on only their group type, either AD or CTRL. We found that all features resulted in a significant difference (p‐value&lt;0.01). Conclusion Though there was no notable visual variation between AD and CTRL spirals, we did find there were significant differences among dynamical characteristics like duration, loop speed, and frequencies. We can conclude that AD participants tend to take longer drawing spirals which result in lower Average Speed per Loop than their CTRL counterparts. Additionally, the FF extracted from both groups indicate that higher frequencies are found in AD participants.
BibTeX
@inproceedings{chavez2024dynamics,
  author = {Gabrielle Chavez and Thomas Thebaud and Ankur Butala and Najim Dehak and Laureano Moro-Velázquez and Esther S. Oh},
  title = {{Dynamics of Handwriting for Cognitive Assessment}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.093190},
}
Figure